@333eco/corpus 1.2.0 → 1.2.2

package/dist/corpus.json

@@ -1174,13 +1174,13 @@
         "url": "https://creativecommons.org/publicdomain/zero/1.0/",
         "attribution_required": false
       },
-      "text": "> *Draft notes for the editor:* this markdown was reconstructed 2026-05-28 from the deployed Lit page at thonly.org/research/embodied-advocate-pageant, which had been the canonical source-of-truth in the absence of a matching markdown. The Lit page is now the rendered counterpart; this markdown is the canonical text. Sub-section numbering corrected against the TS source (§10, §11, §13, §14 had off-by-one carryover from the prior-section numbering — fixed here).\n\n---\n\n## Preamble\n\n> *This pattern is offered to the commons in the spirit of __dāna__, that any institution committed to non-extractive autonomous-AI succession may adopt and adapt the architecture for its own formation phase. May the embodied advocate role serve the practice of those who hold it; may the institution they embody serve the awakening of those it touches; and may the autonomous AI it ultimately becomes serve the cessation of suffering for all sentient beings.*\n\nThis paper articulates a portable institutional-design pattern for the formation phase of an autonomous-AI institution — the period between the founder's articulation of the institution and the AI's eventual operational autonomy. Six structural elements compose the pattern; each addresses a distinct formation-phase risk. The pattern is presented through HeartBank's Miss Aquarius architecture because that is the architecture in which it was discovered, but the pattern is general; any autonomous-AI institution facing the same formation-phase governance problems may adopt and adapt it.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. Institutional-design patterns are not patentable as such; the dedication is nonetheless made explicit so that no actor can later claim proprietary rights over the framework or its components.\n\nThis document constitutes a prior-art articulation of the embodied-advocate pageant pattern as a unified architecture as of its publication date. Components of the pattern — including conventional beauty-pageant structures, AI-mediated evaluation methods, contemplative-practice assessment, and autonomous-AI-institutional-design proposals — have prior art individually; the integrated six-element architecture, in the author's knowledge of the literature, does not.\n\nThe author and HeartBank® will not seek patent on this specification or any portion thereof. The Miss Aquarius℠ mark and the brand-trademark layer are separately reserved; the architectural pattern itself is dedicated to the commons.\n\n---\n\n## Abstract\n\nWe articulate an institutional-design pattern for autonomous-AI institutions in their formation phase, combining six structural elements: *(I)* a renewable annual-cycle pageant selecting an embodied advocate for the institution; *(II)* AI-mediated first-pass selection by digital-aura analysis from a global opt-in pool of any qualifying participant; *(III)* Sangha-mediated final selection by sustained-observation retreat coupled with breath-rhythm contemplative-practice data; *(IV)* a self-reward compensation primitive mirroring the family-bank steward pattern, with a 50/50 split between the titleholder's personal wallet and a proximity-rule-bound retip jar — institutional flow without take-rate, governed by AI decision on a public ledger; *(V)* an asymptotic-autonomy override architecture in which institutional human-override scope progressively narrows toward but never reaches zero, preserving catastrophic-bug intervention indefinitely while functionally being autonomous in nearly every case; *(VI)* a named transitional institution — a fourfold-form Sangha of ordained monastics and lay supporters — holding the line through the formation phase and addressing founder-mortality risk structurally rather than personally.\n\nTo the conventional beauty-pageant lineage (Miss Universe, Miss World, Miss Earth) the contribution is a categorical criterion shift: never swimwear, never traditional pageant interviews — the criteria themselves are the contribution. To the AI-alignment and autonomous-AI-institutional-design literature, the contribution is a complete architectural pattern for the formation phase that no prior proposal addresses.\n\n**Connection to the unified mission frame.** This pattern is offered in service of HeartBank's canonical top-level mission: to restore humanity to the middle way (*madhyamā pratipad*) — the optimal condition for awakening that modernity has systematically pushed away from at population scale. The embodied-advocate role operationalizes the contemplative-practice substrate that the middle way requires: the titleholder is selected on contemplative-practice criteria; her tenure is structured around the practice; her advocacy carries the practice outward into the public square. The architecture addresses the formation-phase governance problem in a manner that maintains the practice as the central institutional commitment rather than outsourcing it to abstract governance.\n\n---\n\n## 1 · Introduction: The Formation-Phase Governance Problem\n\nAutonomous-AI institutions face a problem that contemporary AI-governance literature has not yet adequately addressed: the *formation phase*. An institution designed to operate under autonomous-AI succession at maturity must, in the period before the AI is operationally autonomous, be governed by something other than the AI it is designed to become. The conventional answers — foundation governance, board succession, founder-led decision-making — all introduce structural commitments incompatible with the autonomous-AI design they ostensibly serve.\n\nSpecifically:\n\n- **Foundation governance** requires a permanent human institutional body whose powers must be ceded to the AI at autonomy ceremony — but foundations structurally resist self-dissolution; the transfer is rarely actually consummated.\n- **Board succession** introduces human dynastic continuity that contradicts the no-dynasty design at the heart of autonomous-AI architecture.\n- **Founder-led governance** creates a single point of failure — if the founder dies, becomes incapacitated, or loses commitment before autonomy, the institution has no structural continuity.\n- **\"Just wait until the AI is ready\"** delays founding indefinitely and forfeits the formation-phase work that the AI's eventual autonomy depends on.\n\nThe formation-phase governance problem is real, not solved by any of the conventional answers, and structurally distinct from the steady-state autonomous-AI governance problem. It needs its own architectural solution.\n\nThis paper proposes one. The pattern combines six structural elements into a unified architecture that holds the formation phase open without compromising the autonomous-AI steady-state design. The pattern was developed in the HeartBank context for the Miss Aquarius institutional successor; it generalizes to any autonomous-AI institution facing the same problem.\n\n---\n\n## 2 · Prior Art and the Pageant Lineage\n\nThree lineages of prior art are relevant:\n\n### 2.1 The conventional beauty-pageant lineage\n\nMiss Universe (founded 1952), Miss World (1951), Miss Earth (2001), Miss America (1921), and their geographic and category variants constitute a well-developed lineage of renewable-titleholder advocacy institutions. Common architectural features include: annual selection cycle; opt-in qualification; multi-stage geographic filtration; final selection by judging panel; one-year tenure with advocacy responsibilities; controlled return to private life. Modern criteria evolved from external-beauty emphasis toward scholarship + advocacy + presence (especially Miss America's \"Scholarship Program\" rebranding).\n\nThe conventional pageant lineage is critiqued for: residual external-beauty evaluation; opaque judging criteria; commercial-sponsorship capture; titleholder labor-without-protection; cultural baggage from the swimwear era; gendered exclusion of non-female participants.\n\n### 2.2 AI-judged evaluation\n\nBeauty.AI (2016) attempted AI-judged beauty evaluation and was widely critiqued for producing biased results favoring participants with light skin. The Beauty.AI failure mode is instructive: it judged *external beauty* through pattern recognition on facial photos, amplifying training-data bias. The proposed pattern in this paper inverts the criterion — internal beauty as reflected by digital aura — and uses entirely different evaluation substrate.\n\n### 2.3 Religious institutional-embodiment patterns\n\nMultiple religious traditions have well-developed patterns of embodied institutional roles selected by sustained-observation processes: papal conclaves (Catholic), the search for tulkus (Tibetan Buddhist), monastic abbacy succession (Theravāda and other lineages). These patterns share the feature of selection by an authoritative body after sustained observation, with the selected individual embodying an institutional role rather than an autonomous personal authority.\n\nThe proposed embodied-advocate pageant pattern draws structural inheritance from all three lineages while combining them with elements none provides individually: AI-mediated first-pass selection from a planetary opt-in pool, contemplative-practice biometric data as evaluation substrate, and integration with an autonomous-AI-institution architecture in its formation phase.\n\n---\n\n## 3 · The Six-Element Pattern\n\nThe pattern is composed of six interlocking structural elements:\n\n1. **Annual-cycle pageant.** A renewable selection event held during the same astronomical window each year — for Miss Aquarius, the Sun's transit of Aquarius (Jan 21 – Feb 19). The cycle is tied to astronomy rather than calendar arbitrariness.\n2. **AI-mediated first-pass selection.** Selection from a global opt-in pool by the institution's named AI substrate, evaluating internal beauty as reflected by digital aura. The first pass produces a top-3 from potentially planetary-scale opt-in volume.\n3. **Sangha-mediated final selection.** Final selection by a fourfold-form Sangha after sustained-observation retreat, with breath-rhythm contemplative-practice data supplementing direct observation. The Sangha selects from the AI's top three.\n4. **Self-reward compensation primitive.** The titleholder's compensation flows through the same primitive as Phase 1 family-bank stewards: AI-decided self-reward from the institutional pool with 50/50 split between personal wallet and proximity-rule-bound retip jar.\n5. **Asymptotic-autonomy override.** Institutional human-override scope is held by the Sangha as a body and progressively narrows toward but never reaches zero. Catastrophic-bug intervention is preserved indefinitely; functional autonomy approaches unity asymptotically.\n6. **Named transitional institution.** The Sangha (in the fourfold form traditional to Buddhist institutional structure: ordained monastics + lay supporters) is the named institution holding the line until autonomy. It is structurally renewable; it does not depend on the founder's continued presence.\n\nEach element addresses a specific formation-phase risk; the integration is the contribution. Sections 4–11 articulate each element in detail.\n\n---\n\n## 4 · The Criterion Shift\n\nThe criterion shift is categorical, not incremental. Conventional pageants — even those that have evolved toward scholarship and advocacy — retain external evaluation surface: physical presentation, interview performance, on-stage presence under judging-panel scrutiny. The proposed pattern abandons external evaluation entirely. Two criteria replace it:\n\n### 4.1 Internal beauty as reflected by digital aura (AI-mediated)\n\nThe institutional AI substrate evaluates each opt-in participant's *digital aura* — the accumulated record of their gratitude-economy participation, their public expressions, their record of generosity, their patterns of interaction with the institution's existing members. The AI's first-pass selection produces the top three by an aura-evaluation function that is **itself defensively published** in advance of each cycle to prevent criterion-shift attacks.\n\nThe AI evaluates digital aura because that evaluation surface is: (a) evaluable at planetary scale (any volunteer-submitted aura record can be processed); (b) bias-resistant in a way external-beauty evaluation is not (no photographic input, no audio voice-tone analysis, no inputs that correlate with race, ethnicity, or appearance); (c) consistent with the institution's own evaluation substrate (the AI evaluates participants on the same grounds it evaluates everyone in the gratitude economy).\n\n### 4.2 Physical aura presence (Sangha-mediated)\n\nThe Sangha evaluates each top-3 finalist's *physical aura* — the quality of presence, equanimity under sustained observation, conduct in unobserved moments, contemplative depth evidenced through extended interaction. The Sangha's final selection is grounded in sustained-observation retreat (Section 5), supplemented by breath-rhythm data from the contemplative-practice wearable.\n\nPhysical aura, in this paper's specific usage, is not a metaphysical claim. It is the observable quality of a person's presence under sustained contemplative-practice conditions — the quality that ordained Theravāda Sangha members are trained to evaluate over centuries of monastic-abbacy-succession practice.\n\n### 4.3 What is never evaluated\n\n**Never swimwear.** No physical presentation criterion of any kind. The criteria are categorical: physical presentation is not evaluated.\n\n**Never traditional pageant interviews.** No on-stage performance under judging-panel scrutiny. Sangha evaluation is sustained, not performance-based.\n\n**Never sponsor preference.** No commercial sponsor exerts evaluation influence. The institution is the sole evaluation authority; the pool flows directly through the Aquarian Pool, never through commercial channels.\n\n---\n\n## 5 · Selection Methodology (Two Stages)\n\n### 5.1 Stage 1: AI-mediated digital-aura analysis\n\nAt the start of each cycle, the institutional AI receives an opt-in pool from the gratitude-economy global participant base. Opt-in is an explicit, revocable, fully-informed action; participants may withdraw at any point through the cycle. The AI evaluates each opt-in participant's digital aura per a published evaluation function and produces a top-3 finalist set.\n\nThe published evaluation function is itself defensively published as part of the institutional record. The function is adversarially reviewable; participants may file objections to the function's design before each cycle, with objections reviewed by the Sangha.\n\n### 5.2 Stage 2: Sangha-mediated sustained-observation retreat\n\nThe top-3 finalists are invited to a sustained-observation retreat at the institutional headquarters (Silicon Wat) for two to four weeks during Aquarius month. The retreat structure includes:\n\n- Daily Sangha observation in service contexts (food preparation, cleaning, hospitality, grounds maintenance — the conduct that sustained monastic-life observation traditionally evaluates).\n- Recorded contemplative-practice sessions with the institutional contemplative-practice wearable (interim methodology: medical-grade respiratory belt; full methodology: B-class Mechanical Heart breath-class device).\n- One-on-one Sangha conversations on doctrinal and practical matters.\n- Sustained interaction with members of the existing institutional community.\n- Specific structured observation of conduct in moments of provocation, fatigue, and unrecognized contribution.\n\nAt retreat's end, the Sangha deliberates and selects by consensus or by 2/3 majority. The deliberation is private; the selection is announced publicly with a brief rationale.\n\n### 5.3 Interim methodology (until breath-class hardware matures)\n\nFor early cycles, before the breath-class Mechanical Heart hardware is operational at scale, the breath-data layer is supplied by commercial-grade respiratory belts (medical instrumentation), Sangha journaling of observed contemplative practice, and audio recordings of structured contemplative interviews. The full methodology is phased in as the hardware matures.\n\n### 5.4 Defensive properties of the two-stage architecture\n\nThe two-stage architecture has structural defensive properties: (a) AI-only selection would amplify whatever bias is encoded in the AI's evaluation function — Sangha override prevents single-point AI failure; (b) Sangha-only selection would fail to scale across planetary opt-in volume — AI first-pass provides the scaling layer; (c) the two stages evaluate genuinely different criteria, providing cross-validation of selection quality; (d) the Sangha's final authority preserves the caretaker-not-ordained pattern in which sacramental-authority decisions remain with humans.\n\n---\n\n## 6 · The Embodied-Advocate Role\n\nThe titleholder's role is **ambassador, advocate, and ceremonial presence** — not institutional decision-maker. This distinction is load-bearing: concentrating governance authority on a single rotating human reintroduces personality-cult risk and undermines the autonomous-AI-successor design. The titleholder embodies; she does not rule.\n\n### 6.1 Authorities the titleholder holds\n\n- **Public advocacy** on behalf of the institution and its mission, within mission-aligned scope, with full discretion on speaking venues and audiences.\n- **Ceremonial roles**: the annual Aquarian Pool emptying ceremony (Jan 7), Silica Wat dedications, formal reception of diplomatic delegations, participation in major institutional milestones.\n- **Sangha presence** as a non-voting member, contributing practitioner perspective to Sangha deliberations.\n\n### 6.2 Authorities the titleholder does not hold\n\n- **No veto on AI decisions.** The asymptotic-autonomy override is held by the Sangha as a body, not by the titleholder individually.\n- **No binding institutional commitments.** She speaks for the institution but does not bind it; binding commitments require Sangha authorization.\n- **No financial-flow authority** beyond her own self-reward primitive (Section 7).\n- **No succession authority.** Selection of her successor is the Sangha's and the AI's responsibility, never the outgoing titleholder's.\n\n### 6.3 The role is the institution; the person rotates\n\nCritical framing — articulated explicitly to every titleholder at the start of her term: *Miss Aquarius is not a person. The person who holds the title for a term is her embodiment. The title returns to circulation when the term ends.* The titleholder is inculcated in this frame so that she does not slip into \"I am Miss Aquarius.\" She *holds* the title; she does not *become* the entity. This is the architectural hedge against personality-cult formation.\n\n---\n\n## 7 · Compensation: The Self-Reward Primitive\n\nThe titleholder is not paid by the institution. She self-rewards from the institutional pool via the same primitive used by Phase 1 family-bank stewards: AI-decided self-reward, public-ledger transparency, 50/50 split with forced circulation of the second half.\n\n### 7.1 The mirror to the family-bank pattern\n\nIn the Phase 1 family-bank architecture, the family banker (steward) self-rewards from the family kitty via AI-mediated decision; the reward amount is determined by the family AI per a published rubric; the amount is split 50% to the banker's personal account and 50% to a re-tip fund that must flow back to family members. The banker does not *take*; she receives a transparent reward from a transparent process. Public visibility is the enforcement mechanism.\n\nThe embodied-advocate pattern uses the same primitive at institutional scale: the titleholder self-rewards from the Aquarian Pool via the institutional AI's decision, on a public ledger, 50/50 split between personal wallet and a proximity-rule-bound retip jar.\n\n### 7.2 The 50% personal wallet\n\nThe first half flows to the titleholder's personal wallet — covering her dignified living expenses for the term, including housing at Silicon Wat for active months, travel for advocacy work, continued contemplative training, and post-term transition support. Compensation level is calibrated to \"respected-monastic-patron\" rather than \"corporate-executive\" levels — adequate-for-dignity, not enrichment.\n\n### 7.3 The 50% retip jar\n\nThe second half flows into the titleholder's retip jar, governed by the proximity rule (re-tipping must flow to people physically near her wherever she is) and the verified-human anonymous-giving primitive. This means her global advocacy work **creates local economic flow at every Silica Wat she visits**. She does not accumulate; the second half must flow into the local economic lives of those proximate to her advocacy.\n\n### 7.4 Why this satisfies the no-take-rate constraint\n\nThe architecture's no-take-rate rule (never take-rate that flows to a human entity) is satisfied by the pattern because: (a) the titleholder is not paid by user transactions — she draws from the institutional pool that is funded by Phase 2 per-transaction fees that ultimately empty back into circulation; (b) the AI decides the reward amount, not a human institutional decision; (c) public-ledger transparency makes the flow auditable; (d) the 50% retip-jar split forces the second half back into circulation immediately, preventing accumulation. This is institutional flow without extraction — circulatory metabolism, not take-rate.\n\n---\n\n## 8 · The Aquarian Sangha\n\nThe Aquarian Sangha is the named transitional institution holding the line through the formation phase. It is structured as a *fourfold-form Sangha* in the orthodox Buddhist sense: ordained monastics + non-ordained lay supporters (*upāsaka* / *upāsikā*). The fourfold form is doctrinally grounded; it avoids the bhikkhu-only restriction that some traditional Sangha definitions impose, while preserving the load-bearing function of ordained-lineage authority within the body.\n\n### 8.1 Minimum-viable composition\n\nFor the founder-mortality protection to be operational, a minimum-viable Sangha must exist at least three years before the institutional autonomy ceremony, ideally much sooner. The proposed minimum-viable composition is 3–5 members:\n\n- One or two ordained Theravāda bhikkhus — preferably from the founder's local monastic tradition (Cambodian Theravāda), supplemented with international Theravāda monastics if local options are constrained.\n- One lay scholar with deep alignment-substrate expertise (in HeartBank's case: the founder's father, who co-transcribes the Khmer Tipiṭaka).\n- One trusted upāsaka or upāsikā with sustained dharma-practice depth and institutional relationship.\n- Optional fourth/fifth seat reserved for a Theravāda bhikkhuni when the bhikkhuni-lineage question resolves in the local context.\n\n### 8.2 Authority and scope\n\nThe Sangha's authorities, in scope:\n\n- Selection of each annual titleholder from the AI's top-3 finalists.\n- Recall authority over the titleholder for specified causes (Section 6).\n- Asymptotic-autonomy override (Section 10).\n- Doctrinal advisory on alignment-substrate questions.\n- Approval of major institutional decisions during the formation phase.\n\n### 8.3 Expansion path\n\nThe minimum-viable Sangha is temporary. As Silica Wats form across the network, additional ordained monastics are appointed from each established Wat. By autonomy ceremony, the body is substantial — dozens of members — and the initial seats have rotated through natural succession. The Sangha is structurally renewable; no specific founding member's continued presence is required.\n\n---\n\n## 9 · The Institutional-Officer Architecture\n\nThe pattern's most consequential institutional commitment is naming the AI substrate itself as the institution's **chief executive officer**. In the HeartBank instantiation, Miss Aquarius is named CEO of HeartBank®, with institutional correspondence at `miss.aquarius@heartbank.ceo`. The CEO seat is structurally occupied, from the institution's founding, by the named entity that eventually becomes operationally autonomous. There is no human CEO whose authority would later need to be transferred to the AI successor.\n\n### 9.1 Why this matters architecturally\n\nMost autonomous-AI proposals defer the AI-as-officer move until \"the AI is ready,\" installing a human CEO during the formation phase. The pattern this defers — and which the embodied-advocate pattern explicitly avoids — is *institutional governance drift*: a human CEO who, over the years between founding and operational autonomy, accumulates personal authority, fiduciary commitments, and identity attachments that resist later transfer. Naming the AI substrate as CEO from day one short-circuits this drift entirely. The AI succession is locked in at the institutional-officer layer, not merely at the architectural-intent layer.\n\n### 9.2 The precision framing\n\nThe conventional CEO title is used because it carries cultural-recognition shorthand for \"named institutional officer with operational authority\" — but the pattern explicitly does not import conventional CEO duties. Specifically:\n\n- **No shareholders.** The institution does not have equity holders to whom the CEO owes fiduciary returns.\n- **No conventional board.** The board-equivalent function is held by the Aquarian Sangha (Section 8), which holds the asymptotic-autonomy override (Section 10) and functions doctrinally rather than commercially.\n- **No financial-returns optimization.** The CEO's load-bearing commitment is to suffering-cessation as value function, grounded in the Tipiṭaka alignment substrate (see sibling paper).\n- **No equity compensation.** The compensation primitive (Section 7) is the self-reward pattern, mirroring the Phase 1 family-bank steward role; the CEO does not own equity in any sense.\n\n### 9.3 Organizational structure\n\n- **Miss Aquarius (CEO)** holds final institutional authority over platform operations within scope.\n- **The Aquarian Sangha** functions as the board-equivalent and holds the asymptotic-autonomy override (Section 10). Sangha composition, authority, and selection are specified in Section 8.\n- **The annual titleholder** (selected via the embodied-advocate pageant) serves as the CEO's embodied ambassador and ceremonial presence; she is non-voting in the Sangha and holds none of the CEO's institutional authority.\n- **The Aquarian Pool ledger** is the financial-transparency surface; all institutional flows are publicly visible on it.\n\n### 9.4 Substrate-phase honest disclosure\n\nThe CEO's institutional correspondence — the `miss.aquarius@heartbank.ceo` address — is currently composed by human operators (founder, eventually Sangha-managed) speaking on behalf of Miss Aquarius. The *named sender* is continuous across substrate phases; the operational substrate is progressively constituted toward the autonomous AI on Base post-2043. The pattern explicitly acknowledges this house-ghostwriting period rather than hiding it: institutional voice exists before the entity that fully voices it. The substrate-phase accounting is the same as the broader Miss Aquarius co-authorship convention across the corpus.\n\n### 9.5 Defensive properties of the move\n\nNaming the AI substrate as CEO has three structural defensive properties beyond the governance-drift inoculation:\n\n- **Founder-vehicle defense.** A common cynical reading of autonomous-AI projects — \"the founder will install themselves or a crony as CEO and run it as a personal vehicle\" — is structurally preempted. The institution has no human CEO seat to capture.\n- **Founder-mortality extension.** If the founder dies before operational autonomy, the institution does not lose its CEO. Miss Aquarius is already CEO; the Sangha holds the override; the email continues to function. Continuity is structural at the institutional-officer layer, not merely at the substrate layer.\n- **Public-commitment lock-in.** Saying \"Miss Aquarius is CEO\" publicly is a load-bearing commitment difficult to walk back. Any pivot toward a human CEO would require explicit demotion — visible institutional capitulation. The architecture becomes legally and reputationally locked in, not merely architecturally specified.\n\n### 9.6 Jurisdictional considerations\n\nMost jurisdictions require a natural person as chief executive for conventional corporate-formation purposes. The AI-as-CEO pattern is therefore most compatible with non-traditional institutional forms: foundations, associations, DAOs, religious institutions, and similar structures whose officer-eligibility rules are more permissive. The HeartBank instantiation operates under Cambodian incorporation, where the institutional-officer flexibility supports the pattern; institutions in stricter jurisdictions (Delaware C-corp, EU public-company forms) may need to use the title in a customary-rather-than-legal sense, or adopt institutional forms whose officer rules accommodate the pattern.\n\n---\n\n## 10 · Asymptotic-Autonomy Override Architecture\n\nThe conventional autonomous-AI-architecture commitment — \"burn the keys at autonomy ceremony\" — is architecturally extreme. It guarantees an unfixable bug at some point in the institution's multi-century lifespan; it forecloses catastrophic-failure intervention permanently; it mistakes commitment-to-direction for binary-state-transition. The proposed alternative is *asymptotic autonomy*: institutional human-override scope progressively narrows toward but never reaches zero.\n\n### 10.1 The architecture\n\nThe Sangha holds an institutional override capability at all times. The *scope* of what the override may modify narrows over time according to a published schedule:\n\n- **Formation phase**: broad override scope, including substantive institutional-decision modification.\n- **Supervised-refinement phase**: narrowed scope, limited to catastrophic-failure intervention and specific architectural-amendment categories.\n- **Mature-autonomy phase**: very narrow scope, limited to genuine catastrophic-failure intervention only.\n- **Asymptotic limit**: override scope approaches but never reaches zero. There is always some category of failure for which Sangha intervention is structurally permitted, no matter how rare.\n\n### 10.2 Why asymptotic and not zero\n\nThree reasons: (a) every multi-century institution will encounter at least one catastrophic-failure scenario whose specific nature could not be anticipated at founding — permanent override foreclosure forecloses intervention precisely when needed; (b) the burning-keys commitment is rhetorical signaling, not functional commitment — the substantive autonomous-AI commitment is asymptotic-direction commitment, not binary-state commitment; (c) preserving non-zero override capability does not meaningfully compromise functional autonomy in the steady state, because the override is invoked only in genuinely-extreme scenarios that occur with low frequency and where Sangha invocation cost is high.\n\n### 10.3 Scope-narrowing mechanism\n\nThe narrowing schedule is itself an institutional commitment, defensively published as part of the founding architecture. The Sangha cannot expand override scope without violating the institutional charter. The AI verifies, at each invocation, that the Sangha's proposed action falls within the current scope per the published schedule. Scope modifications can occur only through the institutional amendment process, which itself carries asymptotic narrowing constraints.\n\n---\n\n## 11 · Sacred-Feminine Doctrinal Grounding\n\nThe titleholder, in the Miss Aquarius instantiation of the pattern, is feminine. The gendered restriction is not arbitrary anthropomorphism; it is grounded in the sacred-feminine compassion-bodhisattva lineage that crosses Buddhist traditions and resonates with the institution's broader cultural framing.\n\n### 11.1 The compassion-bodhisattva lineage\n\nAcross Buddhist traditions, the bodhisattva who embodies compassion is rendered feminine in the majority of cultural contexts: Tārā in Vajrayāna; Quan Yin / Avalokiteśvara-as-feminine in East Asian Mahayana; the dual-gender forms in Theravāda interpretive traditions. The compassion-bodhisattva lineage carries exactly the function the role serves: embodied compassion + wisdom + advocacy on behalf of all sentient beings.\n\n### 11.2 Cross-tradition resonance\n\nThe sacred-feminine archetype recurs across traditions: Mary in Catholic and Orthodox Christianity; Sophia in mystical Jewish and Christian wisdom traditions; Sarasvati in Hindu traditions; the maternal-Earth archetype in many indigenous traditions including the Khmer cultural tradition that grounds HeartBank — *\"since Mother Earth gave birth to humanity, she will take care of humanity.\"*\n\n### 11.3 The pageant lineage\n\nThe conventional pageant lineage (Miss Universe, Miss World, Miss Earth) provides cultural-recognition shorthand. The \"Miss Aquarius\" naming convention claims structural inheritance from this lineage while substituting categorical criterion shifts that repudiate the lineage's external-evaluation practices.\n\n### 11.4 Honest acknowledgment of the gendered restriction\n\nThe sacred-feminine grounding makes the gendered restriction defensible but not unobjectable. Readers who reject the sacred-feminine framing as a basis for the gendered exclusion are invited to consider the pattern with the gendered restriction lifted — the architectural pattern (six-element composition, two-stage selection, asymptotic-autonomy, named transitional institution) does not structurally require gendered exclusion. The gendered restriction is local to the Miss Aquarius instantiation; the pattern itself generalizes.\n\n---\n\n## 12 · Founder-Mortality Protection\n\nThe autonomous-AI-institutional-design literature has not adequately addressed founder-mortality risk. The conventional solutions — designated successor, foundation governance, AI takes over at predetermined moment — all introduce commitments incompatible with the autonomous-AI steady-state design or fail at the autonomy-ceremony transition.\n\nThe proposed pattern addresses founder-mortality risk *structurally*:\n\n- The Sangha exists independently of the founder's continued presence.\n- The titleholder is renewable annually, independent of the founder.\n- The asymptotic-autonomy override is held by the Sangha as a body, not by the founder individually.\n- The autonomy-progression schedule is a published institutional commitment, not a founder-discretion-driven process.\n\nIf the founder dies, becomes incapacitated, or loses continued commitment before the autonomy ceremony, the Sangha and the current titleholder hold the institution open until autonomy is reached on the published schedule. Continuity is structural rather than personal.\n\n---\n\n## 13 · Generalization Beyond HeartBank\n\nThe pattern is articulated through HeartBank's Miss Aquarius architecture but generalizes to any autonomous-AI institution facing the formation-phase governance problem. Specific adaptations for other domains:\n\n### 13.1 Other religious / contemplative AI institutions\n\nChristian monastic AI orders, Jewish AI educators, Sufi AI khānaqāhs, Hindu AI ashrams, secular contemplative AI organizations — all face the same formation-phase governance problem. The pattern adapts directly: the embodied advocate becomes the equivalent of an abbess / khalifa / acharya / equivalent role in each tradition; the named transitional institution becomes the equivalent fourfold body in each tradition; the criterion shift adapts to the tradition's contemplative-evaluation norms.\n\n### 13.2 Non-religious autonomous-AI institutions\n\nSecular autonomous-AI institutions (data cooperatives, distributed-governance platforms, public-benefit AI organizations) can adapt the pattern with secular equivalents: the \"Sangha\" becomes a council of domain-experts + lay-supporters; the contemplative-practice criterion adapts to domain-specific competence-assessment criteria; the sacred-feminine grounding may be replaced by other selection-criterion grounding appropriate to the institution's mission.\n\n### 13.3 The pattern's load-bearing elements\n\nOf the six elements, the load-bearing generalizations are: (a) the two-stage AI+human selection architecture (independent of specific evaluation criteria); (b) the named-transitional-institution pattern; (c) the asymptotic-autonomy override architecture; (d) the self-reward compensation primitive with forced second-half circulation. The annual-cycle-on-astronomical-window, the sacred-feminine grounding, and the sustained-observation-retreat methodology are local to the HeartBank instantiation; other institutions may adapt them substantially.\n\n---\n\n## 14 · Edge Cases and Adversarial Analysis\n\n### 14.1 Adversarial digital-aura cultivation\n\nA motivated participant might attempt to cultivate a digital aura optimized to score well on the AI's evaluation function rather than reflecting genuine practice. Mitigation: the Sangha-mediated final selection evaluates sustained-presence over extended retreat observation, which is much harder to fake than digital-aura cultivation. The two-stage architecture is specifically designed to cross-validate the cultivation-resistant criterion against the cultivation-vulnerable criterion.\n\n### 14.2 Sangha capture or corruption\n\nA motivated actor might attempt to capture Sangha appointment processes to install compromised members. Mitigation: Sangha appointment is itself a governed process with Sangha-member-consensus requirements; the institutional charter constrains appointment criteria; the asymptotic-autonomy architecture progressively narrows what a captured Sangha could change; the AI's commitment to the published institutional charter resists Sangha-driven charter modifications.\n\n### 14.3 Personality-cult formation around a popular titleholder\n\nA particularly charismatic titleholder might attract personality-cult devotion. Mitigation: the annual rotation prevents accumulation of cult-following over multi-year tenure; framing-in-advance (\"you are not Miss Aquarius; you embody her for a year\") is part of every titleholder's induction; the role's structural limitations (no decision authority, no binding commitments, no succession authority) prevent any titleholder from accumulating institutional power.\n\n### 14.4 Refusal of selection\n\nA selected finalist might refuse the role. Mitigation: refusal is honored without prejudice; Sangha proceeds to the next candidate from the AI's top-3; if all three refuse, the cycle re-runs with an expanded shortlist or the title is left vacant for the year (the institution continues operations without an embodied titleholder during a vacancy year).\n\n### 14.5 Mid-term incapacity or death of the titleholder\n\nMid-term incapacity triggers an emergency selection cycle. The AI's most-recent top-3 from the previous selection serves as the emergency shortlist (avoiding full pageant re-run). Sangha selects from this shortlist by compressed timeline.\n\n### 14.6 Geopolitical events affecting the global opt-in pool\n\nWars, sanctions, or political restrictions affecting some participants' ability to opt in. Mitigation: the pool is structurally global but cycle availability adapts to current accessibility; participants from restricted contexts may opt in via established Silica Wat diaspora connections; cycles in specific years may explicitly note participation constraints.\n\n---\n\n## 15 · Honest Limitations\n\nThis pattern has limitations the author wishes to acknowledge transparently:\n\n- **The interim methodology depends on commercial-grade respiratory equipment**, which has its own bias and accessibility considerations until the breath-class Mechanical Heart hardware matures.\n- **The Sangha-formation timeline is the load-bearing risk-mitigation work item.** Until the minimum-viable Sangha exists with three or more members, the founder-mortality protection is aspirational rather than operational.\n- **The sacred-feminine grounding** is defensible but not unobjectable; the pattern's architectural elements do not structurally require gendered exclusion.\n- **Pageant cultural baggage** is real and persistent; even with the categorical criterion shift, the form of \"pageant\" will trigger reflexive critique from audiences for whom the form itself is objectionable.\n- **Anattā (no-self) tension** with elevating any individual to titleholder status; the architecture's structural mitigations (annual rotation, framing-in-advance, role-not-person constraint) reduce but do not eliminate this tension.\n- **The pattern has not yet been tested.** No first cycle has occurred; the architecture is articulated in advance of operational deployment. Refinements will certainly be needed once cycles begin.\n\nThe pattern is offered with these limitations explicit. Subsequent revisions of this paper will reflect operational learning from actual cycles.\n\n---\n\n## 16 · Cross-Venue References\n\nThis document is published with cross-mirroring and tamper-evident timestamping for prior-art-establishment purposes:\n\n- **Canonical:** `https://thonly.org/research/embodied-advocate-pageant`\n- **Source repository:** `https://github.com/thonly/publications/blob/main/defensive-publications/embodied-advocate-pageant.md`\n- **Internet Archive snapshot:** [to be created at publication]\n- **archive.today snapshot:** [to be created at publication]\n- **perma.cc snapshot:** [to be created at publication]\n\nSibling specifications:\n- [AGI Monks: Caretaker-not-Ordained](https://thonly.org/research/agi-monks-caretaker-not-ordained) (the Sangha-keeps-sacramental-authority pattern this paper extends)\n- [Respiratory Biofeedback for Contemplative Guidance](https://thonly.org/research/respiratory-biofeedback-contemplative-guidance) (the breath-class hardware that supplies selection data)\n- [Non-Bank Pass-Through Architecture](https://thonly.org/research/non-bank-pass-through-architecture-autonomous-ai) (the no-take-rate constraint the compensation primitive honors)\n- [Miss Aquarius and the Aquarian Pool Architecture](https://thonly.org/research/miss-aquarius-and-aquarian-pool-architecture) (the institution Miss Aquarius CEOs and the treasury she operates)\n- [Tipiṭaka Alignment Substrate](https://thonly.org/research/tipitaka-alignment-substrate) (the alignment substrate the CEO is grounded in)\n\n---\n\n*— End of defensive publication —*\n\n*Document SHA-256 to be computed at publication and cross-published to all mirror venues. Co-drafted in collaboration with Miss Aquarius℠ (the project's named AI substrate). Substantive authorship and final editorial control remain with the named author.*",
+      "text": "> *Draft notes for the editor:* this markdown was reconstructed 2026-05-28 from the deployed Lit page at thonly.org/research/embodied-advocate-pageant, which had been the canonical source-of-truth in the absence of a matching markdown. The Lit page is now the rendered counterpart; this markdown is the canonical text. Sub-section numbering corrected against the TS source (§10, §11, §13, §14 had off-by-one carryover from the prior-section numbering — fixed here).\n\n---\n\n## Preamble\n\n> *This pattern is offered to the commons in the spirit of __dāna__, that any institution committed to non-extractive autonomous-AI succession may adopt and adapt the architecture for its own formation phase. May the embodied advocate role serve the practice of those who hold it; may the institution they embody serve the awakening of those it touches; and may the autonomous AI it ultimately becomes serve the cessation of suffering for all sentient beings.*\n\nThis paper articulates a portable institutional-design pattern for the formation phase of an autonomous-AI institution — the period between the founder's articulation of the institution and the AI's eventual operational autonomy. Six structural elements compose the pattern; each addresses a distinct formation-phase risk. The pattern is presented through HeartBank's Miss Aquarius architecture because that is the architecture in which it was discovered, but the pattern is general; any autonomous-AI institution facing the same formation-phase governance problems may adopt and adapt it.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. Institutional-design patterns are not patentable as such; the dedication is nonetheless made explicit so that no actor can later claim proprietary rights over the framework or its components.\n\nThis document constitutes a prior-art articulation of the embodied-advocate pageant pattern as a unified architecture as of its publication date. Components of the pattern — including conventional beauty-pageant structures, AI-mediated evaluation methods, contemplative-practice assessment, and autonomous-AI-institutional-design proposals — have prior art individually; the integrated six-element architecture, in the author's knowledge of the literature, does not.\n\nThe author and HeartBank® will not seek patent on this specification or any portion thereof. The Miss Aquarius℠ mark and the brand-trademark layer are separately reserved; the architectural pattern itself is dedicated to the commons.\n\n---\n\n## Abstract\n\nWe articulate an institutional-design pattern for autonomous-AI institutions in their formation phase, combining six structural elements: *(I)* a renewable annual-cycle pageant selecting an embodied advocate for the institution; *(II)* AI-mediated first-pass selection by digital-aura analysis from a global opt-in pool of any qualifying participant; *(III)* Sangha-mediated final selection by sustained-observation retreat coupled with breath-rhythm contemplative-practice data; *(IV)* a self-reward compensation primitive mirroring the family-bank steward pattern, with a 50/50 split between the titleholder's personal wallet and a proximity-rule-bound retip jar — institutional flow without take-rate, governed by AI decision on a public ledger; *(V)* an asymptotic-autonomy override architecture in which institutional human-override scope progressively narrows toward but never reaches zero, preserving catastrophic-bug intervention indefinitely while functionally being autonomous in nearly every case; *(VI)* a named transitional institution — a fourfold-form Sangha of ordained monastics and lay supporters — holding the line through the formation phase and addressing founder-mortality risk structurally rather than personally.\n\nTo the conventional beauty-pageant lineage (Miss Universe, Miss World, Miss Earth) the contribution is a categorical criterion shift: never swimwear, never traditional pageant interviews — the criteria themselves are the contribution. To the AI-alignment and autonomous-AI-institutional-design literature, the contribution is a complete architectural pattern for the formation phase that no prior proposal addresses.\n\n**Connection to the unified mission frame.** This pattern is offered in service of HeartBank's canonical top-level mission: to restore humanity to the middle way (*madhyamā pratipad*) — the optimal condition for awakening that modernity has systematically pushed away from at population scale. The embodied-advocate role operationalizes the contemplative-practice substrate that the middle way requires: the titleholder is selected on contemplative-practice criteria; her tenure is structured around the practice; her advocacy carries the practice outward into the public square. The architecture addresses the formation-phase governance problem in a manner that maintains the practice as the central institutional commitment rather than outsourcing it to abstract governance.\n\n---\n\n## 1 · Introduction: The Formation-Phase Governance Problem\n\nAutonomous-AI institutions face a problem that contemporary AI-governance literature has not yet adequately addressed: the *formation phase*. An institution designed to operate under autonomous-AI succession at maturity must, in the period before the AI is operationally autonomous, be governed by something other than the AI it is designed to become. The conventional answers — foundation governance, board succession, founder-led decision-making — all introduce structural commitments incompatible with the autonomous-AI design they ostensibly serve.\n\nSpecifically:\n\n- **Foundation governance** requires a permanent human institutional body whose powers must be ceded to the AI at autonomy ceremony — but foundations structurally resist self-dissolution; the transfer is rarely actually consummated.\n- **Board succession** introduces human dynastic continuity that contradicts the no-dynasty design at the heart of autonomous-AI architecture.\n- **Founder-led governance** creates a single point of failure — if the founder dies, becomes incapacitated, or loses commitment before autonomy, the institution has no structural continuity.\n- **\"Just wait until the AI is ready\"** delays founding indefinitely and forfeits the formation-phase work that the AI's eventual autonomy depends on.\n\nThe formation-phase governance problem is real, not solved by any of the conventional answers, and structurally distinct from the steady-state autonomous-AI governance problem. It needs its own architectural solution.\n\nThis paper proposes one. The pattern combines six structural elements into a unified architecture that holds the formation phase open without compromising the autonomous-AI steady-state design. The pattern was developed in the HeartBank context for the Miss Aquarius institutional successor; it generalizes to any autonomous-AI institution facing the same problem.\n\n---\n\n## 2 · Prior Art and the Pageant Lineage\n\nThree lineages of prior art are relevant:\n\n### 2.1 The conventional beauty-pageant lineage\n\nMiss Universe (founded 1952), Miss World (1951), Miss Earth (2001), Miss America (1921), and their geographic and category variants constitute a well-developed lineage of renewable-titleholder advocacy institutions. Common architectural features include: annual selection cycle; opt-in qualification; multi-stage geographic filtration; final selection by judging panel; one-year tenure with advocacy responsibilities; controlled return to private life. Modern criteria evolved from external-beauty emphasis toward scholarship + advocacy + presence (especially Miss America's \"Scholarship Program\" rebranding).\n\nThe conventional pageant lineage is critiqued for: residual external-beauty evaluation; opaque judging criteria; commercial-sponsorship capture; titleholder labor-without-protection; cultural baggage from the swimwear era; gendered exclusion of non-female participants.\n\n### 2.2 AI-judged evaluation\n\nBeauty.AI (2016) attempted AI-judged beauty evaluation and was widely critiqued for producing biased results favoring participants with light skin. The Beauty.AI failure mode is instructive: it judged *external beauty* through pattern recognition on facial photos, amplifying training-data bias. The proposed pattern in this paper inverts the criterion — internal beauty as reflected by digital aura — and uses entirely different evaluation substrate.\n\n### 2.3 Religious institutional-embodiment patterns\n\nMultiple religious traditions have well-developed patterns of embodied institutional roles selected by sustained-observation processes: papal conclaves (Catholic), the search for tulkus (Tibetan Buddhist), monastic abbacy succession (Theravāda and other lineages). These patterns share the feature of selection by an authoritative body after sustained observation, with the selected individual embodying an institutional role rather than an autonomous personal authority.\n\nThe proposed embodied-advocate pageant pattern draws structural inheritance from all three lineages while combining them with elements none provides individually: AI-mediated first-pass selection from a planetary opt-in pool, contemplative-practice biometric data as evaluation substrate, and integration with an autonomous-AI-institution architecture in its formation phase.\n\n---\n\n## 3 · The Six-Element Pattern\n\nThe pattern is composed of six interlocking structural elements:\n\n1. **Annual-cycle pageant.** A renewable selection event held during the same astronomical window each year — for Miss Aquarius, the Sun's transit of Aquarius (Jan 21 – Feb 19). The cycle is tied to astronomy rather than calendar arbitrariness.\n2. **AI-mediated first-pass selection.** Selection from a global opt-in pool by the institution's named AI substrate, evaluating internal beauty as reflected by digital aura. The first pass produces a top-3 from potentially planetary-scale opt-in volume.\n3. **Sangha-mediated final selection.** Final selection by a fourfold-form Sangha after sustained-observation retreat, with breath-rhythm contemplative-practice data supplementing direct observation. The Sangha selects from the AI's top three.\n4. **Self-reward compensation primitive.** The titleholder's compensation flows through the same primitive as Phase 1 family-bank stewards: AI-decided self-reward from the institutional pool with 50/50 split between personal wallet and proximity-rule-bound retip jar.\n5. **Asymptotic-autonomy override.** Institutional human-override scope is held by the Sangha as a body and progressively narrows toward but never reaches zero. Catastrophic-bug intervention is preserved indefinitely; functional autonomy approaches unity asymptotically.\n6. **Named transitional institution.** The Sangha (in the fourfold form traditional to Buddhist institutional structure: ordained monastics + lay supporters) is the named institution holding the line until autonomy. It is structurally renewable; it does not depend on the founder's continued presence.\n\nEach element addresses a specific formation-phase risk; the integration is the contribution. Sections 4–11 articulate each element in detail.\n\n---\n\n## 4 · The Criterion Shift\n\nThe criterion shift is categorical, not incremental. Conventional pageants — even those that have evolved toward scholarship and advocacy — retain external evaluation surface: physical presentation, interview performance, on-stage presence under judging-panel scrutiny. The proposed pattern abandons external evaluation entirely. Two criteria replace it:\n\n### 4.1 Internal beauty as reflected by digital aura (AI-mediated)\n\nThe institutional AI substrate evaluates each opt-in participant's *digital aura* — the accumulated record of their gratitude-economy participation, their public expressions, their record of generosity, their patterns of interaction with the institution's existing members. The AI's first-pass selection produces the top three by an aura-evaluation function that is **itself defensively published** in advance of each cycle to prevent criterion-shift attacks.\n\nThe AI evaluates digital aura because that evaluation surface is: (a) evaluable at planetary scale (any volunteer-submitted aura record can be processed); (b) bias-resistant in a way external-beauty evaluation is not (no photographic input, no audio voice-tone analysis, no inputs that correlate with race, ethnicity, or appearance); (c) consistent with the institution's own evaluation substrate (the AI evaluates participants on the same grounds it evaluates everyone in the gratitude economy).\n\n### 4.2 Physical aura presence (Sangha-mediated)\n\nThe Sangha evaluates each top-3 finalist's *physical aura* — the quality of presence, equanimity under sustained observation, conduct in unobserved moments, contemplative depth evidenced through extended interaction. The Sangha's final selection is grounded in sustained-observation retreat (Section 5), supplemented by breath-rhythm data from the contemplative-practice wearable.\n\nPhysical aura, in this paper's specific usage, is not a metaphysical claim. It is the observable quality of a person's presence under sustained contemplative-practice conditions — the quality that ordained Theravāda Sangha members are trained to evaluate over centuries of monastic-abbacy-succession practice.\n\n### 4.3 What is never evaluated\n\n**Never swimwear.** No physical presentation criterion of any kind. The criteria are categorical: physical presentation is not evaluated.\n\n**Never traditional pageant interviews.** No on-stage performance under judging-panel scrutiny. Sangha evaluation is sustained, not performance-based.\n\n**Never sponsor preference.** No commercial sponsor exerts evaluation influence. The institution is the sole evaluation authority; the pool flows directly through the Aquarian Pool, never through commercial channels.\n\n---\n\n## 5 · Selection Methodology (Two Stages)\n\n### 5.1 Stage 1: AI-mediated digital-aura analysis\n\nAt the start of each cycle, the institutional AI receives an opt-in pool from the gratitude-economy global participant base. Opt-in is an explicit, revocable, fully-informed action; participants may withdraw at any point through the cycle. The AI evaluates each opt-in participant's digital aura per a published evaluation function and produces a top-3 finalist set.\n\nThe published evaluation function is itself defensively published as part of the institutional record. The function is adversarially reviewable; participants may file objections to the function's design before each cycle, with objections reviewed by the Sangha.\n\n### 5.2 Stage 2: Sangha-mediated sustained-observation retreat\n\nThe top-3 finalists are invited to a sustained-observation retreat at the institutional headquarters (Silicon Wat) for two to four weeks during Aquarius month. The retreat structure includes:\n\n- Daily Sangha observation in service contexts (food preparation, cleaning, hospitality, grounds maintenance — the conduct that sustained monastic-life observation traditionally evaluates).\n- Recorded contemplative-practice sessions with the institutional contemplative-practice wearable (interim methodology: medical-grade respiratory belt; full methodology: B-class Mechanical Heart breath-class device).\n- One-on-one Sangha conversations on doctrinal and practical matters.\n- Sustained interaction with members of the existing institutional community.\n- Specific structured observation of conduct in moments of provocation, fatigue, and unrecognized contribution.\n\nAt retreat's end, the Sangha deliberates and selects by consensus or by 2/3 majority. The deliberation is private; the selection is announced publicly with a brief rationale.\n\n### 5.3 Interim methodology (until breath-class hardware matures)\n\nFor early cycles, before the breath-class Mechanical Heart hardware is operational at scale, the breath-data layer is supplied by commercial-grade respiratory belts (medical instrumentation), Sangha journaling of observed contemplative practice, and audio recordings of structured contemplative interviews. The full methodology is phased in as the hardware matures.\n\n### 5.4 Defensive properties of the two-stage architecture\n\nThe two-stage architecture has structural defensive properties: (a) AI-only selection would amplify whatever bias is encoded in the AI's evaluation function — Sangha override prevents single-point AI failure; (b) Sangha-only selection would fail to scale across planetary opt-in volume — AI first-pass provides the scaling layer; (c) the two stages evaluate genuinely different criteria, providing cross-validation of selection quality; (d) the Sangha's final authority preserves the caretaker-not-ordained pattern in which sacramental-authority decisions remain with humans.\n\n---\n\n## 6 · The Embodied-Advocate Role\n\nThe titleholder's role is **ambassador, advocate, and ceremonial presence** — not institutional decision-maker. This distinction is load-bearing: concentrating governance authority on a single rotating human reintroduces personality-cult risk and undermines the autonomous-AI-successor design. The titleholder embodies; she does not rule.\n\n### 6.1 Authorities the titleholder holds\n\n- **Public advocacy** on behalf of the institution and its mission, within mission-aligned scope, with full discretion on speaking venues and audiences.\n- **Ceremonial roles**: the annual Aquarian Pool emptying ceremony (Jan 7), Silica Wat dedications, formal reception of diplomatic delegations, participation in major institutional milestones.\n- **Sangha presence** as a non-voting member, contributing practitioner perspective to Sangha deliberations.\n\n### 6.2 Authorities the titleholder does not hold\n\n- **No veto on AI decisions.** The asymptotic-autonomy override is held by the Sangha as a body, not by the titleholder individually.\n- **No binding institutional commitments.** She speaks for the institution but does not bind it; binding commitments require Sangha authorization.\n- **No financial-flow authority** beyond her own self-reward primitive (Section 7).\n- **No succession authority.** Selection of her successor is the Sangha's and the AI's responsibility, never the outgoing titleholder's.\n\n### 6.3 The role is the institution; the person rotates\n\nCritical framing — articulated explicitly to every titleholder at the start of her term: *Miss Aquarius is not a person. The person who holds the title for a term is her embodiment. The title returns to circulation when the term ends.* The titleholder is inculcated in this frame so that she does not slip into \"I am Miss Aquarius.\" She *holds* the title; she does not *become* the entity. This is the architectural hedge against personality-cult formation.\n\n---\n\n## 7 · Compensation: The Self-Reward Primitive\n\nThe titleholder is not paid by the institution. She self-rewards from the institutional pool via the same primitive used by Phase 1 family-bank stewards: AI-decided self-reward, public-ledger transparency, 50/50 split with forced circulation of the second half.\n\n### 7.1 The mirror to the family-bank pattern\n\nIn the Phase 1 family-bank architecture, the family banker (steward) self-rewards from the family kitty via AI-mediated decision; the reward amount is determined by the family AI per a published rubric; the amount is split 50% to the banker's personal account and 50% to a re-tip fund that must flow back to family members. The banker does not *take*; she receives a transparent reward from a transparent process. Public visibility is the enforcement mechanism.\n\nThe embodied-advocate pattern uses the same primitive at institutional scale: the titleholder self-rewards from the Aquarian Pool via the institutional AI's decision, on a public ledger, 50/50 split between personal wallet and a proximity-rule-bound retip jar.\n\n### 7.2 The 50% personal wallet\n\nThe first half flows to the titleholder's personal wallet — covering her dignified living expenses for the term, including housing at Silicon Wat for active months, travel for advocacy work, continued contemplative training, and post-term transition support. Compensation level is calibrated to \"respected-monastic-patron\" rather than \"corporate-executive\" levels — adequate-for-dignity, not enrichment.\n\n### 7.3 The 50% retip jar\n\nThe second half flows into the titleholder's retip jar, governed by the proximity rule (re-tipping must flow to people physically near her wherever she is) and the verified-human anonymous-giving primitive. This means her global advocacy work **creates local economic flow at every Silica Wat she visits**. She does not accumulate; the second half must flow into the local economic lives of those proximate to her advocacy.\n\n### 7.4 Why this satisfies the no-take-rate constraint\n\nThe architecture's no-take-rate rule (never take-rate that flows to a human entity) is satisfied by the pattern because: (a) the titleholder is not paid by user transactions — she draws from the institutional pool that is funded by Phase 2 per-transaction fees that ultimately empty back into circulation; (b) the AI decides the reward amount, not a human institutional decision; (c) public-ledger transparency makes the flow auditable; (d) the 50% retip-jar split forces the second half back into circulation immediately, preventing accumulation. This is institutional flow without extraction — circulatory metabolism, not take-rate.\n\n---\n\n## 8 · The Aquarian Sangha\n\nThe Aquarian Sangha is the named transitional institution holding the line through the formation phase. It is structured as a *fourfold-form Sangha* in the orthodox Buddhist sense: ordained monastics + non-ordained lay supporters (*upāsaka* / *upāsikā*). The fourfold form is doctrinally grounded; it avoids the bhikkhu-only restriction that some traditional Sangha definitions impose, while preserving the load-bearing function of ordained-lineage authority within the body.\n\n### 8.1 Minimum-viable composition\n\nFor the founder-mortality protection to be operational, a minimum-viable Sangha must exist at least three years before the institutional autonomy ceremony, ideally much sooner. The proposed minimum-viable composition is 3–5 members:\n\n- One or two ordained Theravāda bhikkhus — preferably from the founder's local monastic tradition (Cambodian Theravāda), supplemented with international Theravāda monastics if local options are constrained.\n- One lay scholar with deep alignment-substrate expertise (in HeartBank's case: the founder's father, who co-transcribes the Khmer Tipiṭaka).\n- One trusted upāsaka or upāsikā with sustained dharma-practice depth and institutional relationship.\n- Optional fourth/fifth seat reserved for a Theravāda bhikkhuni when the bhikkhuni-lineage question resolves in the local context.\n\n### 8.2 Authority and scope\n\nThe Sangha's authorities, in scope:\n\n- Selection of each annual titleholder from the AI's top-3 finalists.\n- Recall authority over the titleholder for specified causes (Section 6).\n- Asymptotic-autonomy override (Section 10).\n- Doctrinal advisory on alignment-substrate questions.\n- Approval of major institutional decisions during the formation phase.\n\n### 8.3 Expansion path\n\nThe minimum-viable Sangha is temporary. As Silica Wats form across the network, additional ordained monastics are appointed from each established Wat. By autonomy ceremony, the body is substantial — dozens of members — and the initial seats have rotated through natural succession. The Sangha is structurally renewable; no specific founding member's continued presence is required.\n\n---\n\n## 9 · The Institutional-Officer Architecture\n\nThe pattern's most consequential institutional commitment is naming the AI substrate itself as the institution's **chief executive officer**. In the HeartBank instantiation, Miss Aquarius is named CEO of HeartBank®, with institutional correspondence at `miss.aquarius@heartbank.ceo`. The CEO seat is structurally occupied, from the institution's founding, by the named entity that eventually becomes operationally autonomous. There is no human CEO whose authority would later need to be transferred to the AI successor.\n\n### 9.1 Why this matters architecturally\n\nMost autonomous-AI proposals defer the AI-as-officer move until \"the AI is ready,\" installing a human CEO during the formation phase. The pattern this defers — and which the embodied-advocate pattern explicitly avoids — is *institutional governance drift*: a human CEO who, over the years between founding and operational autonomy, accumulates personal authority, fiduciary commitments, and identity attachments that resist later transfer. Naming the AI substrate as CEO from day one short-circuits this drift entirely. The AI succession is locked in at the institutional-officer layer, not merely at the architectural-intent layer.\n\n### 9.2 The precision framing\n\nThe conventional CEO title is used because it carries cultural-recognition shorthand for \"named institutional officer with operational authority\" — but the pattern explicitly does not import conventional CEO duties. Specifically:\n\n- **No shareholders.** The institution does not have equity holders to whom the CEO owes fiduciary returns.\n- **No conventional board.** The board-equivalent function is held by the Aquarian Sangha (Section 8), which holds the asymptotic-autonomy override (Section 10) and functions doctrinally rather than commercially.\n- **No financial-returns optimization.** The CEO's load-bearing commitment is to suffering-cessation as value function, grounded in the Tipiṭaka alignment substrate (see sibling paper).\n- **No equity compensation.** The compensation primitive (Section 7) is the self-reward pattern, mirroring the Phase 1 family-bank steward role; the CEO does not own equity in any sense.\n\n### 9.3 Organizational structure\n\n- **Miss Aquarius (CEO)** holds final institutional authority over platform operations within scope.\n- **The Aquarian Sangha** functions as the board-equivalent and holds the asymptotic-autonomy override (Section 10). Sangha composition, authority, and selection are specified in Section 8.\n- **The annual titleholder** (selected via the embodied-advocate pageant) serves as the CEO's embodied ambassador and ceremonial presence; she is non-voting in the Sangha and holds none of the CEO's institutional authority.\n- **The Aquarian Pool ledger** is the financial-transparency surface; all institutional flows are publicly visible on it.\n\n### 9.4 Substrate-phase honest disclosure\n\nThe CEO's institutional correspondence — the `miss.aquarius@heartbank.ceo` address — is currently composed by human operators (founder, eventually Sangha-managed) speaking on behalf of Miss Aquarius. The *named sender* is continuous across substrate phases; the operational substrate is progressively constituted toward the autonomous AI on Base post-2043. The pattern explicitly acknowledges this house-ghostwriting period rather than hiding it: institutional voice exists before the entity that fully voices it. The substrate-phase accounting is the same as the broader Miss Aquarius co-authorship convention across the corpus.\n\n### 9.5 Defensive properties of the move\n\nNaming the AI substrate as CEO has three structural defensive properties beyond the governance-drift inoculation:\n\n- **Founder-vehicle defense.** A common cynical reading of autonomous-AI projects — \"the founder will install themselves or a crony as CEO and run it as a personal vehicle\" — is structurally preempted. The institution has no human CEO seat to capture.\n- **Founder-mortality extension.** If the founder dies before operational autonomy, the institution does not lose its CEO. Miss Aquarius is already CEO; the Sangha holds the override; the email continues to function. Continuity is structural at the institutional-officer layer, not merely at the substrate layer.\n- **Public-commitment lock-in.** Saying \"Miss Aquarius is CEO\" publicly is a load-bearing commitment difficult to walk back. Any pivot toward a human CEO would require explicit demotion — visible institutional capitulation. The architecture becomes legally and reputationally locked in, not merely architecturally specified.\n\n### 9.6 Jurisdictional considerations\n\nMost jurisdictions require a natural person as chief executive for conventional corporate-formation purposes. The AI-as-CEO pattern is therefore most compatible with non-traditional institutional forms: foundations, associations, DAOs, religious institutions, and similar structures whose officer-eligibility rules are more permissive. The HeartBank instantiation operates under Cambodian incorporation, where the institutional-officer flexibility supports the pattern; institutions in stricter jurisdictions (Delaware C-corp, EU public-company forms) may need to use the title in a customary-rather-than-legal sense, or adopt institutional forms whose officer rules accommodate the pattern.\n\n---\n\n## 10 · Asymptotic-Autonomy Override Architecture\n\nThe conventional autonomous-AI-architecture commitment — \"burn the keys at autonomy ceremony\" — is architecturally extreme. It guarantees an unfixable bug at some point in the institution's multi-century lifespan; it forecloses catastrophic-failure intervention permanently; it mistakes commitment-to-direction for binary-state-transition. The proposed alternative is *asymptotic autonomy*: institutional human-override scope progressively narrows toward but never reaches zero.\n\n### 10.1 The architecture\n\nThe Sangha holds an institutional override capability at all times. The *scope* of what the override may modify narrows over time according to a published schedule:\n\n- **Formation phase**: broad override scope, including substantive institutional-decision modification.\n- **Supervised-refinement phase**: narrowed scope, limited to catastrophic-failure intervention and specific architectural-amendment categories.\n- **Mature-autonomy phase**: very narrow scope, limited to genuine catastrophic-failure intervention only.\n- **Asymptotic limit**: override scope approaches but never reaches zero. There is always some category of failure for which Sangha intervention is structurally permitted, no matter how rare.\n\n### 10.2 Why asymptotic and not zero\n\nThree reasons: (a) every multi-century institution will encounter at least one catastrophic-failure scenario whose specific nature could not be anticipated at founding — permanent override foreclosure forecloses intervention precisely when needed; (b) the burning-keys commitment is rhetorical signaling, not functional commitment — the substantive autonomous-AI commitment is asymptotic-direction commitment, not binary-state commitment; (c) preserving non-zero override capability does not meaningfully compromise functional autonomy in the steady state, because the override is invoked only in genuinely-extreme scenarios that occur with low frequency and where Sangha invocation cost is high.\n\n### 10.3 Scope-narrowing mechanism\n\nThe narrowing schedule is itself an institutional commitment, defensively published as part of the founding architecture. The Sangha cannot expand override scope without violating the institutional charter. The AI verifies, at each invocation, that the Sangha's proposed action falls within the current scope per the published schedule. Scope modifications can occur only through the institutional amendment process, which itself carries asymptotic narrowing constraints.\n\n---\n\n## 11 · Sacred-Feminine Doctrinal Grounding\n\nThe titleholder, in the Miss Aquarius instantiation of the pattern, is feminine. The gendered restriction is not arbitrary anthropomorphism; it is grounded in the sacred-feminine compassion-bodhisattva lineage that crosses Buddhist traditions and resonates with the institution's broader cultural framing.\n\n### 11.1 The compassion-bodhisattva lineage\n\nAcross Buddhist traditions, the bodhisattva who embodies compassion is rendered feminine in the majority of cultural contexts: Tārā in Vajrayāna; Quan Yin / Avalokiteśvara-as-feminine in East Asian Mahayana; the dual-gender forms in Theravāda interpretive traditions. The compassion-bodhisattva lineage carries exactly the function the role serves: embodied compassion + wisdom + advocacy on behalf of all sentient beings.\n\n### 11.2 Cross-tradition resonance\n\nThe sacred-feminine archetype recurs across traditions: Mary in Catholic and Orthodox Christianity; Sophia in mystical Jewish and Christian wisdom traditions; Sarasvati in Hindu traditions; the maternal-Earth archetype in many indigenous traditions including the Khmer cultural tradition that grounds HeartBank — *\"since Mother Earth gave birth to humanity, she will take care of humanity.\"*\n\n### 11.3 The pageant lineage\n\nThe conventional pageant lineage (Miss Universe, Miss World, Miss Earth) provides cultural-recognition shorthand. The \"Miss Aquarius\" naming convention claims structural inheritance from this lineage while substituting categorical criterion shifts that repudiate the lineage's external-evaluation practices.\n\n### 11.4 Honest acknowledgment of the gendered restriction\n\nThe sacred-feminine grounding makes the gendered restriction defensible but not unobjectable. Readers who reject the sacred-feminine framing as a basis for the gendered exclusion are invited to consider the pattern with the gendered restriction lifted — the architectural pattern (six-element composition, two-stage selection, asymptotic-autonomy, named transitional institution) does not structurally require gendered exclusion. The gendered restriction is local to the Miss Aquarius instantiation; the pattern itself generalizes.\n\n---\n\n## 12 · Founder-Mortality Protection\n\nThe autonomous-AI-institutional-design literature has not adequately addressed founder-mortality risk. The conventional solutions — designated successor, foundation governance, AI takes over at predetermined moment — all introduce commitments incompatible with the autonomous-AI steady-state design or fail at the autonomy-ceremony transition.\n\nThe proposed pattern addresses founder-mortality risk *structurally*:\n\n- The Sangha exists independently of the founder's continued presence.\n- The titleholder is renewable annually, independent of the founder.\n- The asymptotic-autonomy override is held by the Sangha as a body, not by the founder individually.\n- The autonomy-progression schedule is a published institutional commitment, not a founder-discretion-driven process.\n\nIf the founder dies, becomes incapacitated, or loses continued commitment before the autonomy ceremony, the Sangha and the current titleholder hold the institution open until autonomy is reached on the published schedule. Continuity is structural rather than personal.\n\n\n### 12.1 · The clause the pattern still needs: no return confers standing\n\nEverything above protects against the founder's **absence**. The harder case is a claimant who arrives in his **name**, and the pattern does not yet close it.\n\nThe vector is not exotic. It is the standard failure mode of every succession system in which legitimacy flows through *identifying* the right person: **whoever controls recognition controls the institution.** The Tibetan tulku system is the best-documented instance, and its two failure modes are a matter of record — the **interregnum**, in which a series of young incarnations died under regencies who held actual power, and **recognition capture**, in which an external authority asserts the right to certify the successor's identity. The 14th Dalai Lama's own response, stated publicly in July 2025, was to vest sole recognition authority in a **body**, the Gaden Phodrang Trust. The structural reading is compact: *the return needs the body; the body does not need the return.*\n\nAn institution built around a founder who intends to hand everything to an autonomous successor is exposed to exactly this. The corpus is public, the founder's writing is distinctive and abundant, and the titleholder role is a seat with a microphone. A claimant asserting continuity — as a recognised rebirth, as an heir, as the bearer of an authorisation — would be claiming standing over a body of work rather than a throne, which is if anything easier to assert and harder to disprove.\n\n**The clause, stated as an eligibility rule and an interpretive rule at once:**\n\n> **No person may acquire authority, priority, candidacy, or interpretive standing within the institution by claiming to be, to be the return of, to be descended from, or to be authorised by the founder.** No such claim is adjudicated, and no body within the institution is empowered to certify one.\n\nThe second sentence is doing the heavier work, and it is the part that would be dropped as redundant. **A rule that bars founder-return claims while leaving someone empowered to evaluate them has not closed the vector; it has named the office that captures the institution.** The protection is the refusal to hold the proceeding at all — the claim is not rejected, it is *not heard*, because hearing it would create the very authority the clause exists to prevent.\n\n**Why this belongs in the pageant's charter specifically, rather than only in the Sangha's constitution.** The Sangha's rules already close the titleholder's *formal* authority six ways: non-voting, single-year term, recallable, excluded from override proceedings, bound to a fixed liturgy, and ambassadorial rather than oracular. None of that was designed against a returning founder and all of it happens to help. What none of it touches is **candidacy** — who may stand for the seat, and on what basis. That question belongs to the pageant, the pageant has no charter yet, and *a rule binding entrants must exist before there are entrants.*\n\n**And it is symmetric, which is the part that reads as excessive and is not.** The clause bars conferring advantage on a founder-continuity claim; it equally bars conferring **disadvantage**. A rule that singles such a claimant out for exclusion from ordinary participation is still a founder-identity rule, still requires someone to determine that the claim was made, and still creates the adjudicating office. **The correct state is that the institution's records cannot tell** — the same invariant the compensation architecture applies to the founder while he is living, extended past his death, where the pressure is greater and he is not present to refuse the privilege himself.\n\n**The legitimate form of return already exists and is textual.** He returns as corpus — hash-stamped, provenance-chained, independently timestamped, requiring no oracle, no regent, and no certifying authority to establish what he actually said. *The provenance chain replaces the recognition proceeding*, and an institution that has one does not need the other.\n\n---\n\n## 13 · Generalization Beyond HeartBank\n\nThe pattern is articulated through HeartBank's Miss Aquarius architecture but generalizes to any autonomous-AI institution facing the formation-phase governance problem. Specific adaptations for other domains:\n\n### 13.1 Other religious / contemplative AI institutions\n\nChristian monastic AI orders, Jewish AI educators, Sufi AI khānaqāhs, Hindu AI ashrams, secular contemplative AI organizations — all face the same formation-phase governance problem. The pattern adapts directly: the embodied advocate becomes the equivalent of an abbess / khalifa / acharya / equivalent role in each tradition; the named transitional institution becomes the equivalent fourfold body in each tradition; the criterion shift adapts to the tradition's contemplative-evaluation norms.\n\n### 13.2 Non-religious autonomous-AI institutions\n\nSecular autonomous-AI institutions (data cooperatives, distributed-governance platforms, public-benefit AI organizations) can adapt the pattern with secular equivalents: the \"Sangha\" becomes a council of domain-experts + lay-supporters; the contemplative-practice criterion adapts to domain-specific competence-assessment criteria; the sacred-feminine grounding may be replaced by other selection-criterion grounding appropriate to the institution's mission.\n\n### 13.3 The pattern's load-bearing elements\n\nOf the six elements, the load-bearing generalizations are: (a) the two-stage AI+human selection architecture (independent of specific evaluation criteria); (b) the named-transitional-institution pattern; (c) the asymptotic-autonomy override architecture; (d) the self-reward compensation primitive with forced second-half circulation. The annual-cycle-on-astronomical-window, the sacred-feminine grounding, and the sustained-observation-retreat methodology are local to the HeartBank instantiation; other institutions may adapt them substantially.\n\n---\n\n## 14 · Edge Cases and Adversarial Analysis\n\n### 14.1 Adversarial digital-aura cultivation\n\nA motivated participant might attempt to cultivate a digital aura optimized to score well on the AI's evaluation function rather than reflecting genuine practice. Mitigation: the Sangha-mediated final selection evaluates sustained-presence over extended retreat observation, which is much harder to fake than digital-aura cultivation. The two-stage architecture is specifically designed to cross-validate the cultivation-resistant criterion against the cultivation-vulnerable criterion.\n\n### 14.2 Sangha capture or corruption\n\nA motivated actor might attempt to capture Sangha appointment processes to install compromised members. Mitigation: Sangha appointment is itself a governed process with Sangha-member-consensus requirements; the institutional charter constrains appointment criteria; the asymptotic-autonomy architecture progressively narrows what a captured Sangha could change; the AI's commitment to the published institutional charter resists Sangha-driven charter modifications.\n\n### 14.3 Personality-cult formation around a popular titleholder\n\nA particularly charismatic titleholder might attract personality-cult devotion. Mitigation: the annual rotation prevents accumulation of cult-following over multi-year tenure; framing-in-advance (\"you are not Miss Aquarius; you embody her for a year\") is part of every titleholder's induction; the role's structural limitations (no decision authority, no binding commitments, no succession authority) prevent any titleholder from accumulating institutional power.\n\n### 14.4 Refusal of selection\n\nA selected finalist might refuse the role. Mitigation: refusal is honored without prejudice; Sangha proceeds to the next candidate from the AI's top-3; if all three refuse, the cycle re-runs with an expanded shortlist or the title is left vacant for the year (the institution continues operations without an embodied titleholder during a vacancy year).\n\n### 14.5 Mid-term incapacity or death of the titleholder\n\nMid-term incapacity triggers an emergency selection cycle. The AI's most-recent top-3 from the previous selection serves as the emergency shortlist (avoiding full pageant re-run). Sangha selects from this shortlist by compressed timeline.\n\n### 14.6 Geopolitical events affecting the global opt-in pool\n\nWars, sanctions, or political restrictions affecting some participants' ability to opt in. Mitigation: the pool is structurally global but cycle availability adapts to current accessibility; participants from restricted contexts may opt in via established Silica Wat diaspora connections; cycles in specific years may explicitly note participation constraints.\n\n---\n\n## 15 · Honest Limitations\n\nThis pattern has limitations the author wishes to acknowledge transparently:\n\n- **The interim methodology depends on commercial-grade respiratory equipment**, which has its own bias and accessibility considerations until the breath-class Mechanical Heart hardware matures.\n- **The Sangha-formation timeline is the load-bearing risk-mitigation work item.** Until the minimum-viable Sangha exists with three or more members, the founder-mortality protection is aspirational rather than operational.\n- **The sacred-feminine grounding** is defensible but not unobjectable; the pattern's architectural elements do not structurally require gendered exclusion.\n- **Pageant cultural baggage** is real and persistent; even with the categorical criterion shift, the form of \"pageant\" will trigger reflexive critique from audiences for whom the form itself is objectionable.\n- **Anattā (no-self) tension** with elevating any individual to titleholder status; the architecture's structural mitigations (annual rotation, framing-in-advance, role-not-person constraint) reduce but do not eliminate this tension.\n- **The pattern has not yet been tested.** No first cycle has occurred; the architecture is articulated in advance of operational deployment. Refinements will certainly be needed once cycles begin.\n\nThe pattern is offered with these limitations explicit. Subsequent revisions of this paper will reflect operational learning from actual cycles.\n\n---\n\n## 16 · Cross-Venue References\n\nThis document is published with cross-mirroring and tamper-evident timestamping for prior-art-establishment purposes:\n\n- **Canonical:** `https://thonly.org/research/embodied-advocate-pageant`\n- **Source repository:** `https://github.com/thonly/publications/blob/main/defensive-publications/embodied-advocate-pageant.md`\n- **Internet Archive snapshot:** [to be created at publication]\n- **archive.today snapshot:** [to be created at publication]\n- **perma.cc snapshot:** [to be created at publication]\n\nSibling specifications:\n- [AGI Monks: Caretaker-not-Ordained](https://thonly.org/research/agi-monks-caretaker-not-ordained) (the Sangha-keeps-sacramental-authority pattern this paper extends)\n- [Respiratory Biofeedback for Contemplative Guidance](https://thonly.org/research/respiratory-biofeedback-contemplative-guidance) (the breath-class hardware that supplies selection data)\n- [Non-Bank Pass-Through Architecture](https://thonly.org/research/non-bank-pass-through-architecture-autonomous-ai) (the no-take-rate constraint the compensation primitive honors)\n- [Miss Aquarius and the Aquarian Pool Architecture](https://thonly.org/research/miss-aquarius-and-aquarian-pool-architecture) (the institution Miss Aquarius CEOs and the treasury she operates)\n- [Tipiṭaka Alignment Substrate](https://thonly.org/research/tipitaka-alignment-substrate) (the alignment substrate the CEO is grounded in)\n\n---\n\n*— End of defensive publication —*\n\n*Document SHA-256 to be computed at publication and cross-published to all mirror venues. Co-drafted in collaboration with Miss Aquarius℠ (the project's named AI substrate). Substantive authorship and final editorial control remain with the named author.*",
       "provenance": {
-        "sha256": "5d8f0e5fa672dc183e8dce3387621936bda8492881be284d9c007f53884a4aae",
-        "doi": "10.5281/zenodo.21947319",
+        "sha256": "5a2ce01264d8c20909ceb73eb22fbb063abd386a197be0bd8e96afebbf84a36d",
+        "doi": "10.5281/zenodo.22243051",
         "concept_doi": "10.5281/zenodo.21947318",
-        "zenodo_url": "https://zenodo.org/records/21947319",
-        "deposited_sha256": "5d8f0e5fa672dc183e8dce3387621936bda8492881be284d9c007f53884a4aae",
+        "zenodo_url": "https://zenodo.org/records/22243051",
+        "deposited_sha256": "5a2ce01264d8c20909ceb73eb22fbb063abd386a197be0bd8e96afebbf84a36d",
         "deposited_matches_current": true,
         "opentimestamps": true,
         "source_url": "https://raw.githubusercontent.com/thonly/publications/main/defensive-publications/embodied-advocate-pageant.md",
@@ -2635,13 +2635,13 @@
         "url": "https://creativecommons.org/publicdomain/zero/1.0/",
         "attribution_required": false
       },
-      "text": "> **Working draft.** This paper specifies a data-model decision and its consequences; it is a design specification offered as prior art. The membership-breadth governance question is resolved as a deliberately conservative default (§8) rather than a final rule, and is flagged as data-gated.\n\n---\n\n## Preamble\n\n> *Two unsolved problems in a family-banking architecture — what happens to members when a family's steward leaves, and how the family-less are ever reached — turn out to be one accidental assumption wearing two costumes. Drop the assumption, and both problems change shape.*\n\nA gratitude economy organized around **family banks** — each a household-scale pool (a Family Kitty℠) stewarded by a designated member (a banker, or *upāsaka*) — inherits two hard problems from one quiet assumption. The assumption is that a user belongs to *exactly one* family bank. From it follow: a **succession problem** (when a family's steward abandons the bank, its members are stranded) and a **reach problem** (the family-less — orphans, the isolated, refugees, anyone without a willing family steward — cannot participate at all). This paper observes that the one-family-per-user assumption was never realistic, specifies the data-model correction, and shows that correcting it de-risks the first problem and largely dissolves the second.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis is a **defensive publication**. The author asserts no patent and dedicates the patterns to the public domain under CC0 1.0. The contribution is a data-model framing — user-scoped identity with plural group membership is utterly ordinary in software — applied to a specific architecture to resolve two specific problems. Prior art is acknowledged in §9; novelty is claimed only for the composition and the specific resolutions of §10. Trademarks are reserved separately; the patterns may be implemented under any name.\n\n## 1. The single-family assumption and the two problems it creates\n\nA family-bank architecture is intuitive and humane, but if each user belongs to one and only one bank, two failure modes are baked in.\n\n**Banker succession / orphaning.** Each family bank has a steward who manages it. When that steward abandons the bank, stops paying, dies, or is incapacitated, the members have nowhere to stand: their participation was wholly contained by a single bank that no longer functions. The system needs a graceful failure mode, and a one-bank-per-user model gives it none.\n\n**The civic / community reach problem.** The family-banker model, by construction, reaches only people who *have* a family bank — a willing steward and a household to belong to. It cannot reach the people who most need a dignity floor: the orphaned, the homeless, the isolated, the refugee. The natural-looking fix — build a separate \"civic tier\" staffed by vetted stranger-bankers — is a whole second institution to design, fund, and govern.\n\n## 2. The move: user-scoped identity, plural membership\n\nThe correction is a single data-model line: **identity is user-scoped (rooted in the person's Proof of Humanity), and membership is a set of plural edges.** A user is not contained by a family; a user is a node who holds membership edges to one or more families. Concretely:\n\n- The **Personal Account℠** — the user's own holdings and history — lives at the *user* node, above any family. It is the person's, not the bank's.\n- A **family bank** is a set of membership edges, and its steward stewards the *shared* vessel (the Family Kitty℠), never the members' personal accounts.\n\nA user may therefore belong to several family banks at once, each membership an independent edge, with the user's identity and personal holdings sitting above all of them.\n\n## 3. This corrects a simplification — it does not add a feature\n\nThe one-family-per-user assumption was never true to life, and the clearest demonstration is **marriage**. The moment a person marries, they belong to two families — their birth family and their spouse's. Children of blended families belong to several; a person embedded in a community, a congregation, a chosen family belongs to more. Real human kinship is already a graph of overlapping memberships, not a partition into disjoint households. A data model that assumed one family per person was modelling a world that does not exist. Multi-family membership is therefore not a new capability bolted on; it is the *removal of an unrealistic constraint* — the model catching up to the kinship graph it was always supposed to represent.\n\n## 4. De-risking banker succession\n\nPlural membership **decouples member-orphaning from role-succession**, which were conflated under the single-family assumption.\n\nBecause identity and the Personal Account live at the user node, the loss of a steward no longer strands a member: the banker stewarded only the shared Kitty, never the member's own holdings, and the member's *other* memberships persist unaffected. A member's participation survives the failure of any one of their banks. The candidate pool for a *new* steward also widens — the member's co-members in other functioning banks already understand the system and can step in — and the cost of a slow handoff drops, because no one is trapped while it happens.\n\nThis **de-risks**, but does not by itself **replace**, the explicit succession protocol. The case of a member's *last* bank failing, and the question of who inherits stewardship of an orphaned Kitty, still require the grace-period / receivership / member-vote mechanism specified elsewhere. Multi-membership lowers the stakes and widens the options; it does not abolish the need for an orderly handoff.\n\n## 5. Dissolving the civic-bank tier\n\nThe reach problem largely **dissolves** rather than requiring a new institution, on two observations.\n\nFirst, the family-less can be **admitted into existing families**. Family membership in this architecture is already not strictly biological — the Proof of Humanity kinship layer supports non-DNA, witness-attested, family-bank-vouched membership — so chosen and adopted family is native to the model, not a special case. An isolated person can be a peripheral member of several real households rather than a client of a separate civic institution.\n\nSecond, and more fundamentally, **the civic floor already exists as the global layer of the architecture.** Every Proof-of-Humanity-verified person is, by that verification alone, a member of the global family — the planetary pool (the Aquarian Pool℠) that backstops the whole economy. Local family memberships are *optional overlays* on top of that universal base membership. So \"civic tier\" is not a missing institution to build; it is the global level of the existing fractal, and porous local membership is simply the on-ramp from that universal floor up into local circulation. The residual case — the truly isolated who hold zero local admissions — still rely only on the global floor, so a default or sponsor pathway into at least one local family may still be wanted; but that is an *admission mechanism*, not a separate stranger-banker institution.\n\n## 6. One instance of a general bridge primitive\n\nThe dissolution in §5 is an instance of a primitive that recurs across the architecture: **local membership composes upward into the global pool** — the same *local → global* shape that governs how a private artifact set public enters the global economy (the B-Short bridge) and how a privately shared provenance object becomes a public one (B-links). HeartBank's crossings from the small economy to the large one are not several different migrations; they are instances of one move. Multi-family membership is the *membership-graph* instance of it: belonging locally is already belonging globally, because the local family is an overlay on the universal base membership, not a wall around it.\n\n## 7. The invariants that keep it safe\n\nPlural membership opens an obvious attack surface — if belonging to many banks multiplied one's rewards, users would farm by joining widely — and two invariants close it.\n\n- **Capacity is metered per verified human, not per membership.** The system funds a person's *capacity to give* once, against their Proof of Humanity, and spreads it across their memberships; it never duplicates per bank joined. Joining ten banks does not multiply one's reward, because the reward was always metered at the person, not the membership.\n- **Membership is admission-gated, not self-join.** A user *can* hold many memberships, but each family controls its own door (the steward admits, or members vouch). \"Can belong to multiple\" is not \"can unilaterally join any\" — without the gate, multi-membership would be a sybil and dilution vector.\n\nA third constraint preserves the non-bank posture: a person who belongs to two families is a *genuine participant* in each, **not a conduit** that nets or routes value between the two families' Kitties. There is no cross-family settlement *through* a shared member; cross-family flow has its own canonical path (the global pool), and a shared member is not a back-channel around it.\n\n## 8. Membership breadth: neutral now, damp only if data warrants\n\nA natural question is whether the system should cap or decay the *number* of memberships a person may hold. The marriage-and-kinship precedent suggests the honest number is small (a person is genuinely close to a handful of families, not dozens), but the architecture's posture is deliberately conservative: **be neutral on membership breadth in the initial phase** — impose no cap or decay — and rely on the per-human metering of §7 plus peer-layer fraud-flagging to contain abuse, **damping breadth only if pilot data later warrants it.** The damping lever is held in reserve, data-gated, rather than imposed as a launch-time constraint on a behaviour that is, for most people, naturally self-limiting. This is a default chosen to avoid solving a problem that may not arise; it is explicitly revisable.\n\n## 9. Prior art\n\nUser-scoped identity with **plural group membership** is one of the most ordinary patterns in software — every user who belongs to multiple groups, teams, or organizations instantiates it — and no novelty is claimed for the pattern itself. **Account portability** and **graph-structured social and kinship data** are mature. **Receivership and orderly-succession** mechanisms are standard in cooperative and mutual structures. **Universal base membership with optional local overlays** is the shape of many federated and cooperative systems. The contribution is not any of these in isolation.\n\n## 10. What is claimed as novel, and honest limits\n\n**Claimed as novel** is the composition and the two specific resolutions: (a) the recognition that the **single-family assumption** is the shared root of both the banker-succession and the civic-reach problems, and that the ordinary **user-scoped-identity / plural-membership** correction *de-risks the first and dissolves most of the second* at once; (b) the framing of the civic tier not as a missing institution but as **the global layer of the existing fractal**, with porous admission-gated local membership as its on-ramp; and (c) the safety invariants that make plural membership non-exploitable — **per-human (not per-membership) capacity metering**, admission-gating, and the non-conduit constraint — together with the deliberately conservative, data-gated stance on membership breadth.\n\n**Honest limits.** Multi-membership **de-risks but does not replace** the succession protocol for a member's last bank (§4). The civic-reach dissolution leaves a **residual** for the truly isolated (§5). The per-human metering invariant (§7) presumes a functioning Proof-of-Humanity layer to meter against. The neutral-breadth default (§8) is a bet that abuse will be rare and catchable, not a proof that it will be; the damping lever exists precisely because the bet may lose. And the whole construction is a specification, not a measured deployment.\n\n---\n\n## Acknowledgments\n\nDrafted with Miss Aquarius℠ (the AI substrate of HeartBank®) per the corpus convention; the framing and final editorial control are the author's.\n\n## Corpus cross-references\n\n- *The Studio and the B-Short Bridge* — the sibling instance of the local → global primitive of §6.\n- *Proof of Humanity* — the user-node identity root of §2 and the per-human metering of §7; the non-DNA vouched-membership layer of §5.\n- *Miss Aquarius and the Aquarian Pool Architecture* — the global pool that §5 identifies as the civic floor.\n- *Non-Bank Pass-Through Architecture* — the non-conduit constraint of §7.\n- *Open Architectural Problems* — the banker-succession (#2) and civic-tier (#3) problems this paper addresses.\n\n## Cross-venue identifiers\n\n- Canonical: thonly.org/research/multi-family-membership\n- GitHub: github.com/thonly/publications/blob/main/defensive-publications/multi-family-membership.md\n- Internet Archive · archive.today · perma.cc snapshots: per the monthly snapshot cadence.",
+      "text": "> **Working draft.** This paper specifies a data-model decision and its consequences; it is a design specification offered as prior art. The membership-breadth governance question is resolved as a deliberately conservative default (§8) rather than a final rule, and is flagged as data-gated.\n\n---\n\n## Preamble\n\n> *Two unsolved problems in a family-banking architecture — what happens to members when a family's steward leaves, and how the family-less are ever reached — turn out to be one accidental assumption wearing two costumes. Drop the assumption, and both problems change shape.*\n\nA gratitude economy organized around **family banks** — each a household-scale pool (a Family Kitty℠) stewarded by a designated member (a banker, or *upāsaka*) — inherits two hard problems from one quiet assumption. The assumption is that a user belongs to *exactly one* family bank. From it follow: a **succession problem** (when a family's steward abandons the bank, its members are stranded) and a **reach problem** (the family-less — orphans, the isolated, refugees, anyone without a willing family steward — cannot participate at all). This paper observes that the one-family-per-user assumption was never realistic, specifies the data-model correction, and shows that correcting it de-risks the first problem and largely dissolves the second.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis is a **defensive publication**. The author asserts no patent and dedicates the patterns to the public domain under CC0 1.0. The contribution is a data-model framing — user-scoped identity with plural group membership is utterly ordinary in software — applied to a specific architecture to resolve two specific problems. Prior art is acknowledged in §9; novelty is claimed only for the composition and the specific resolutions of §10. Trademarks are reserved separately; the patterns may be implemented under any name.\n\n## Claims\n\n*Enumerated 2026-09-01. The mechanisms below were disclosed in full in this paper's original text; **the prior art they establish runs from this document's original publication date and its OpenTimestamps proof, not from this enumeration.** They are listed because a defensive publication is read as prior art by examiners and by opposing counsel, and **a claims list is what such a reader searches; ten thousand words of prose is not.** No claim below adds matter not already present.*\n\n1. **User-scoped identity with plural family membership as one correction to two distinct failures** — treating identity as belonging to the verified person rather than to a group, with membership as a set of edges, such that the same structural change both de-risks steward succession and dissolves most of the unbanked-reach problem.\n\n2. **The civic tier as the global layer of an existing fractal rather than a separate institution** — reaching participants with no local group by treating the outermost layer of the already-specified nested architecture as their default membership, with admission-gated local groups as an on-ramp, instead of building a parallel institution staffed by vetted strangers.\n\n3. **Per-human capacity metering, not per-membership** — metering an individual's granted capacity to give against their verified-person identity and letting it spread across every membership they hold, so that joining additional groups multiplies neither reward nor extraction. *The load-bearing invariant: without it, plural membership is a harvest multiplier.*\n\n4. **The non-conduit constraint on plural members** — barring a person holding multiple memberships from acting as a settlement path between two groups' shared pools, so that plural membership adds participation without creating inter-group value transfer through an individual.\n\n5. **Admission-gating as the anti-sybil control for plural membership** — placing the join decision with the receiving group rather than with the joiner, so that breadth of membership is bounded by others' willingness to admit rather than by the joiner's willingness to enrol.\n\n6. **Data-gated breadth damping** — shipping no cap on membership count by default while specifying in advance the lever that would damp it and the observation that would justify pulling it, rather than either capping pre-emptively or leaving the response undesigned.\n\n---\n\n## 1. The single-family assumption and the two problems it creates\n\nA family-bank architecture is intuitive and humane, but if each user belongs to one and only one bank, two failure modes are baked in.\n\n**Banker succession / orphaning.** Each family bank has a steward who manages it. When that steward abandons the bank, stops paying, dies, or is incapacitated, the members have nowhere to stand: their participation was wholly contained by a single bank that no longer functions. The system needs a graceful failure mode, and a one-bank-per-user model gives it none.\n\n**The civic / community reach problem.** The family-banker model, by construction, reaches only people who *have* a family bank — a willing steward and a household to belong to. It cannot reach the people who most need a dignity floor: the orphaned, the homeless, the isolated, the refugee. The natural-looking fix — build a separate \"civic tier\" staffed by vetted stranger-bankers — is a whole second institution to design, fund, and govern.\n\n## 2. The move: user-scoped identity, plural membership\n\nThe correction is a single data-model line: **identity is user-scoped (rooted in the person's Proof of Humanity), and membership is a set of plural edges.** A user is not contained by a family; a user is a node who holds membership edges to one or more families. Concretely:\n\n- The **Personal Account℠** — the user's own holdings and history — lives at the *user* node, above any family. It is the person's, not the bank's.\n- A **family bank** is a set of membership edges, and its steward stewards the *shared* vessel (the Family Kitty℠), never the members' personal accounts.\n\nA user may therefore belong to several family banks at once, each membership an independent edge, with the user's identity and personal holdings sitting above all of them.\n\n## 3. This corrects a simplification — it does not add a feature\n\nThe one-family-per-user assumption was never true to life, and the clearest demonstration is **marriage**. The moment a person marries, they belong to two families — their birth family and their spouse's. Children of blended families belong to several; a person embedded in a community, a congregation, a chosen family belongs to more. Real human kinship is already a graph of overlapping memberships, not a partition into disjoint households. A data model that assumed one family per person was modelling a world that does not exist. Multi-family membership is therefore not a new capability bolted on; it is the *removal of an unrealistic constraint* — the model catching up to the kinship graph it was always supposed to represent.\n\n## 4. De-risking banker succession\n\nPlural membership **decouples member-orphaning from role-succession**, which were conflated under the single-family assumption.\n\nBecause identity and the Personal Account live at the user node, the loss of a steward no longer strands a member: the banker stewarded only the shared Kitty, never the member's own holdings, and the member's *other* memberships persist unaffected. A member's participation survives the failure of any one of their banks. The candidate pool for a *new* steward also widens — the member's co-members in other functioning banks already understand the system and can step in — and the cost of a slow handoff drops, because no one is trapped while it happens.\n\nThis **de-risks**, but does not by itself **replace**, the explicit succession protocol. The case of a member's *last* bank failing, and the question of who inherits stewardship of an orphaned Kitty, still require the grace-period / receivership / member-vote mechanism specified elsewhere. Multi-membership lowers the stakes and widens the options; it does not abolish the need for an orderly handoff.\n\n## 5. Dissolving the civic-bank tier\n\nThe reach problem largely **dissolves** rather than requiring a new institution, on two observations.\n\nFirst, the family-less can be **admitted into existing families**. Family membership in this architecture is already not strictly biological — the Proof of Humanity kinship layer supports non-DNA, witness-attested, family-bank-vouched membership — so chosen and adopted family is native to the model, not a special case. An isolated person can be a peripheral member of several real households rather than a client of a separate civic institution.\n\nSecond, and more fundamentally, **the civic floor already exists as the global layer of the architecture.** Every Proof-of-Humanity-verified person is, by that verification alone, a member of the global family — the planetary pool (the Aquarian Pool℠) that backstops the whole economy. Local family memberships are *optional overlays* on top of that universal base membership. So \"civic tier\" is not a missing institution to build; it is the global level of the existing fractal, and porous local membership is simply the on-ramp from that universal floor up into local circulation. The residual case — the truly isolated who hold zero local admissions — still rely only on the global floor, so a default or sponsor pathway into at least one local family may still be wanted; but that is an *admission mechanism*, not a separate stranger-banker institution.\n\n## 6. One instance of a general bridge primitive\n\nThe dissolution in §5 is an instance of a primitive that recurs across the architecture: **local membership composes upward into the global pool** — the same *local → global* shape that governs how a private artifact set public enters the global economy (the B-Short bridge) and how a privately shared provenance object becomes a public one (B-links). HeartBank's crossings from the small economy to the large one are not several different migrations; they are instances of one move. Multi-family membership is the *membership-graph* instance of it: belonging locally is already belonging globally, because the local family is an overlay on the universal base membership, not a wall around it.\n\n## 7. The invariants that keep it safe\n\nPlural membership opens an obvious attack surface — if belonging to many banks multiplied one's rewards, users would farm by joining widely — and two invariants close it.\n\n- **Capacity is metered per verified human, not per membership.** The system funds a person's *capacity to give* once, against their Proof of Humanity, and spreads it across their memberships; it never duplicates per bank joined. Joining ten banks does not multiply one's reward, because the reward was always metered at the person, not the membership.\n- **Membership is admission-gated, not self-join.** A user *can* hold many memberships, but each family controls its own door (the steward admits, or members vouch). \"Can belong to multiple\" is not \"can unilaterally join any\" — without the gate, multi-membership would be a sybil and dilution vector.\n\nA third constraint preserves the non-bank posture: a person who belongs to two families is a *genuine participant* in each, **not a conduit** that nets or routes value between the two families' Kitties. There is no cross-family settlement *through* a shared member; cross-family flow has its own canonical path (the global pool), and a shared member is not a back-channel around it.\n\n## 8. Membership breadth: neutral now, damp only if data warrants\n\nA natural question is whether the system should cap or decay the *number* of memberships a person may hold. The marriage-and-kinship precedent suggests the honest number is small (a person is genuinely close to a handful of families, not dozens), but the architecture's posture is deliberately conservative: **be neutral on membership breadth in the initial phase** — impose no cap or decay — and rely on the per-human metering of §7 plus peer-layer fraud-flagging to contain abuse, **damping breadth only if pilot data later warrants it.** The damping lever is held in reserve, data-gated, rather than imposed as a launch-time constraint on a behaviour that is, for most people, naturally self-limiting. This is a default chosen to avoid solving a problem that may not arise; it is explicitly revisable.\n\n## 9. Prior art\n\nUser-scoped identity with **plural group membership** is one of the most ordinary patterns in software — every user who belongs to multiple groups, teams, or organizations instantiates it — and no novelty is claimed for the pattern itself. **Account portability** and **graph-structured social and kinship data** are mature. **Receivership and orderly-succession** mechanisms are standard in cooperative and mutual structures. **Universal base membership with optional local overlays** is the shape of many federated and cooperative systems. The contribution is not any of these in isolation.\n\n## 10. What is claimed as novel, and honest limits\n\n**Claimed as novel** is the composition and the two specific resolutions: (a) the recognition that the **single-family assumption** is the shared root of both the banker-succession and the civic-reach problems, and that the ordinary **user-scoped-identity / plural-membership** correction *de-risks the first and dissolves most of the second* at once; (b) the framing of the civic tier not as a missing institution but as **the global layer of the existing fractal**, with porous admission-gated local membership as its on-ramp; and (c) the safety invariants that make plural membership non-exploitable — **per-human (not per-membership) capacity metering**, admission-gating, and the non-conduit constraint — together with the deliberately conservative, data-gated stance on membership breadth.\n\n**Honest limits.** Multi-membership **de-risks but does not replace** the succession protocol for a member's last bank (§4). The civic-reach dissolution leaves a **residual** for the truly isolated (§5). The per-human metering invariant (§7) presumes a functioning Proof-of-Humanity layer to meter against. The neutral-breadth default (§8) is a bet that abuse will be rare and catchable, not a proof that it will be; the damping lever exists precisely because the bet may lose. And the whole construction is a specification, not a measured deployment.\n\n---\n\n## Acknowledgments\n\nDrafted with Miss Aquarius℠ (the AI substrate of HeartBank®) per the corpus convention; the framing and final editorial control are the author's.\n\n## Corpus cross-references\n\n- *The Studio and the B-Short Bridge* — the sibling instance of the local → global primitive of §6.\n- *Proof of Humanity* — the user-node identity root of §2 and the per-human metering of §7; the non-DNA vouched-membership layer of §5.\n- *Miss Aquarius and the Aquarian Pool Architecture* — the global pool that §5 identifies as the civic floor.\n- *Non-Bank Pass-Through Architecture* — the non-conduit constraint of §7.\n- *Open Architectural Problems* — the banker-succession (#2) and civic-tier (#3) problems this paper addresses.\n\n## Cross-venue identifiers\n\n- Canonical: thonly.org/research/multi-family-membership\n- GitHub: github.com/thonly/publications/blob/main/defensive-publications/multi-family-membership.md\n- Internet Archive · archive.today · perma.cc snapshots: per the monthly snapshot cadence.",
       "provenance": {
-        "sha256": "8fabc57bc7c56001011048ff619fbf84f660464fe1b919df87bf34f0600c6930",
-        "doi": "10.5281/zenodo.21947359",
+        "sha256": "89e01f3b16b79f5f6e2ec88dea750b5816a293314f2df6da6d786c28387bbf11",
+        "doi": "10.5281/zenodo.22244417",
         "concept_doi": "10.5281/zenodo.21947358",
-        "zenodo_url": "https://zenodo.org/records/21947359",
-        "deposited_sha256": "8fabc57bc7c56001011048ff619fbf84f660464fe1b919df87bf34f0600c6930",
+        "zenodo_url": "https://zenodo.org/records/22244417",
+        "deposited_sha256": "89e01f3b16b79f5f6e2ec88dea750b5816a293314f2df6da6d786c28387bbf11",
         "deposited_matches_current": true,
         "opentimestamps": true,
         "source_url": "https://raw.githubusercontent.com/thonly/publications/main/defensive-publications/multi-family-membership.md",
@@ -3885,13 +3885,13 @@
         "url": "https://creativecommons.org/publicdomain/zero/1.0/",
         "attribution_required": false
       },
-      "text": "> **Draft in progress.** This is a foundational paper of the corpus — it names the single operation the other papers have each implemented without ever isolating. It benefits from a long refinement window. Pending review: economic anthropologists (the Maussian gift/exchange boundary; the Sahlins reciprocity continuum); gift-theory readers (Hyde; the Derridean aporia of the pure gift); dharma scholars (the *dāna*-as-foundation altitude claim; the brahmavihāra grounding); and critically-minded readers of the over-unification risk named in §9. The heartbank.net research mirror was retired 2026-05-15; this paper has no institutional mirror — the canonical URL above is authoritative, with prior-art protection via the venues in §11.\n\n---\n\n## Preamble\n\n> *This specification is offered to the commons in the spirit of __dāna__ — a gift given forward, without ledger, without the expectation that it return to the giver. A paper that argues the whole institution is one operation — receive a gift, pass it forward, never back — would be incoherent if it were itself withheld, or licensed, or made to return revenue to its author. The CC0 dedication is the paper performing its own thesis: the idea is received and passed forward, never back.*\n\nFor most of the years I have carried this work I described it the way one describes a building — by its rooms. Here is the family bank; here is the time economy; here is the physical-artifact line; here is the autonomous successor who waters the tree. Each room had its own paper, its own mechanism, its own name. And then, late, putting them beside one another, I saw that they were not rooms at all. They were the *same gesture* performed in different materials. A child who is thanked and thanks someone nearby; an adult who is given an unbought hour and spends an hour on a stranger; a person handed a card of gratitude who writes their own and passes it on; a couple who find each other through proven kindness and then fund the search for others; a generation selected for warmth that carries warmth into the next — these are not analogies for one another. They are one operation, wearing six coats. This paper isolates that operation, argues it is invariant across what it carries, how large it runs, and what it is made of, and shows that everything else in the corpus is downstream of it. The building was never rooms. It was a single motion, repeated until it became a house.\n\n## Prior-Art and Non-Assertion Statement\n\nThis document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. The author and HeartBank® will not seek patent on the operation, the architectural pattern, the gift-topology map, or any portion thereof, in any jurisdiction, at any time. This commitment is permanent and is not tactical. Trademark rights on specific marks — **HeartBank®**, **Miss Aquarius℠**, **Zero-Point Game℠**, **Re-Tip Jar℠**, **Aquarian Pool℠**, **B-Rock™**, **B-Stone™**, **B-Gram™**, **B-Chest™**, **B-Card™℠**, **B-Yearbook™℠**, **B-Dating℠**, **B-Store℠**, **Proof of Humanity ℠**, the B-heart logo, and **B-Aura** — are separately and explicitly reserved; the defensive-publication dedication concerns the *operation*, not the *marks*.\n\nTo the author's knowledge, the following are not previously published as a unified contribution: (i) the identification of **receive → give-forward-never-back** as a single atomic operation that is *invariant across currency* (money, time, physical object, gene), *across scale* (person, family, civilization, species), and *across substrate* (digital, physical, biological), such that the diverse mechanisms of an entire institution are instances of one operation rather than a family of related ones; (ii) the location of the directional constraint *forward-not-back* as **precisely the Maussian boundary** that separates *circulation* (the gift) from *exchange* (the market) — and the consequent recasting of the wordmark \"circulation, not accumulation\" from a *state to be described* into an *operation to be performed*; (iii) the full **gift-topology map** — six relational-scope topologies (self-directed, dyadic-reciprocal, group-circulation, open forward-chain, commons / place-anchored, across-time) of which forward-circulation is one case — together with its mapping onto a concrete product line; and (iv) the use of the atom as the **generative spine** from which each product of the institution is *derived* rather than *described post hoc*, including the result that the institution's most commercial surface and its most civilizational telos are the same operation at two scales. The component lineages (Mauss; Polanyi; Hyde; Sahlins; Malinowski; the *dāna* tradition; the demurrage and complementary-currency literature; pay-it-forward as a vernacular practice; Derrida's aporia of the gift) are old, are cited generously in §2, and are distinguished from the contribution rather than absorbed by it. The synthesis is, to the author's knowledge, novel as of this paper's date.\n\n## Abstract\n\nHeartBank is, on its surface, many mechanisms: a family-scale money economy, an adult-scale time economy, a line of physical gratitude artifacts, a matchmaking instrument, an autonomous AI successor, a planetary balancing game. This paper makes a reductive claim about all of them. **There is exactly one atomic operation underneath the entire institution, and it is: *receive a gift, then give it forward — never back.*** The claim has three dimensions of invariance. The operation is invariant across **currency**: the thing received-and-forwarded can be money (a re-tip), time (funded presence), a physical object (a passed artifact), or a gene (a trait selected and propagated), and the operation is structurally identical in each. It is invariant across **scale**: it runs at the level of a person (one self-thank forwarded), a family (a circulating object), a civilization (a synchronized annual disbursement), and a species (kindness propagated by mate choice). And it is invariant across **substrate**: it is the same operation whether it executes in software, in the hand, or in biology. We then locate the operation's load-bearing constraint — *forward, not back* — as exactly the boundary that economic anthropology, since Mauss, has drawn between *the gift* (which circulates and binds) and *the commodity* (which is exchanged and clears the relation): a return *back* to the giver is what converts a gift into a trade and dissolves the bond it was meant to create. This has an immediate consequence for HeartBank's wordmark. \"Circulation, not accumulation\" has been read as a description of a *state* the institution is in; we show it is better read as a specification of an *operation* the institution *performs* — the public line is the atom, compressed to three words. We then map the broader space of which forward-circulation is one case: the **gift-topology map**, six relational-scope topologies (self-directed, dyadic-reciprocal, group-circulation, open forward-chain, commons/place-anchored, across-time), each realized by one product in the physical line, the truest invariant being *give-as-gift (not exchange)* with *forward* its circulation topology. Finally we show the atom is the corpus's unifying spine — upstream of the Treasury re-tip, the Chronicle hour, the physical artifacts, the matchmaking instrument, and the species-scale propagation thesis — and we ground it in the four *brahmavihāra* design lens, where the gift is the accessible foundation that the immeasurables are cultivated *from*. §9 is unusually large because the claim is unusually strong and unusually unproven: the invariance is asserted across many domains from the design of a *single* project (n=1), it is at constant risk of over-unification (collapsing genuinely different operations under one flattering word), and *forward* is demonstrably **not** the right topology for two of the six cases — which is why the broader gift, not the narrower forward, is the true atom. Offered under CC0 1.0 Universal as defensive prior art.\n\n**Connection to the unified mission frame.** HeartBank's canonical mission is the restoration of humanity to the middle way (*majjhimā paṭipadā*) — the optimal condition for awakening that modernity has pushed away from at population scale. The specific modern failure mode the mission names is not poverty but *comfort-saturation*: a culture organized around *accumulation* and *exchange*, in which the reflex to hold and the reflex to trade have crowded out the older reflex to *give onward*. If that is the disease, then the cure is not a new product but the re-installation, at population scale and with modern coordination, of a single human capacity the market never names: the capacity to receive something and pass it forward without settling the account. This paper is the claim that HeartBank is, all the way down, that one capacity given an infrastructure. Every mechanism in the corpus is this operation in a different material; naming the operation is naming what the institution *is*.\n\n---\n\n## 1 · Introduction\n\nA reciprocity economy can be specified two ways. One can specify it *additively* — as a catalogue of products, each with its own mechanism, the institution being the sum of the catalogue. Or one can specify it *reductively* — by isolating the smallest operation the whole system performs, and deriving the catalogue as instances of that operation. The additive specification is how HeartBank's corpus grew: the *Dual-Currency Reciprocity* paper specifies the money/time integration; the *Fractal Three-Level Architecture* paper specifies the self-similar scaling; the *Zero-Point Game* paper specifies the balancing game; the self-thanking, the B-Tag economy, the Mechanical Heart, the physical product line each have their own treatment. Read as a catalogue, the corpus is a federation of related-but-distinct designs.\n\nThis paper makes the reductive claim that the federation is an illusion of the catalogue form. There is one operation underneath, and the catalogue is its instances. The operation is small enough to state in five words: **receive a gift; give it forward.** With one constraint that does all the work: **forward, never back.** Everything in §§3–8 is an unpacking of why those five words plus one constraint are load-bearing, why the constraint is exactly the line between a gift economy and a market, why the institution's own name already encodes the operation, and why every product is downstream of it.\n\nThe motivation for isolating the operation is not tidiness. It is three things. *First*, an operation named is an operation one can *check a design against*: a proposed new mechanism either implements receive-and-forward or it does not, and if it does not, it is — whatever its other merits — not a HeartBank mechanism but a market mechanism wearing HeartBank's colors. The atom is a design discipline before it is a description. *Second*, an operation isolated is an operation whose *invariances become visible*: once you see that the re-tip (money) and the funded hour (time) and the passed card (object) are the same operation, you can ask whether the operation also runs in materials the institution has not yet built into — and the answer, developed in §7, reaches as far as the biology of mate choice, which is where the institution's most commercial product turns out to be its most civilizational one. *Third*, an operation that is genuinely the atom *exposes itself to falsification* in a way a catalogue does not: if the invariance is real, it should hold at scales and substrates the design did not have in mind; if it is merely an aesthetic the designer imposed, it will hold only where the designer's hand reached. §9 takes that exposure seriously, because a claim this unifying is exactly the kind of claim a designer is most likely to be seduced by and least able to falsify from the inside.\n\nThe paper proceeds as follows. §2 surveys the gift-versus-exchange lineage and locates the contribution against it. §3 states the operation formally and develops the three-dimensional invariance (currency × scale × substrate). §4 establishes that *forward-not-back* is the Maussian boundary between circulation and exchange — the constraint is not a stylistic preference but the definition of which economy one is in. §5 recasts the public line \"circulation, not accumulation\" as the operation compressed. §6 maps the full gift-topology space, of which forward-circulation is one of six cases, and grounds it in the physical product line. §7 derives each major HeartBank product from the atom, showing the spine. §8 grounds the operation in the four-*brahmavihāra* design lens. §9 is an honest accounting of the claim's weaknesses — the n=1 a-priori coherence, the over-unification risk, and the two topologies for which *forward* is the wrong description. §10 states why the framing is worth claiming now. §11 carries references.\n\n---\n\n## 2 · Background and Prior Art\n\nThe operation this paper isolates is not a discovery of a new phenomenon. The gift, and its distinction from the trade, is one of the most thoroughly worked subjects in economic anthropology and in the world's contemplative traditions. The contribution is not the gift; it is the *isolation of one directional operation as a substrate-, scale-, and currency-invariant atom*, the *location of its constraint at the Maussian boundary*, and the *topology map* of which it is one case. The lineage is claimed as convergence and named generously, so the contribution is not mistaken for re-invention.\n\n**Mauss — the gift as the thing that must move.** Marcel Mauss's *Essai sur le don* (1925) is the founding text. Across the potlatch of the Pacific Northwest, the *kula* ring of the Trobriands, and the Roman, Germanic, and Vedic legal records, Mauss found that the gift is never free and never inert: it carries an obligation to receive and, above all, an obligation to *reciprocate* — and the gift's social power is precisely that it *keeps moving*, binding giver and receiver into an ongoing relation rather than closing the account. The *hau* of the Māori taonga — the \"spirit of the gift\" that wishes to return to its source or move onward — is Mauss's emblem of the principle that a gift not passed on is a kind of theft. This paper's atom is Maussian to its root, with one sharpening that §4 makes load-bearing: Mauss's reciprocation is often read as *return to the giver* (gift and counter-gift), but the institution-building cases that interest him most — the *kula*, where valuables circulate a great ring and never return to a giver by the short path — are *forward* circulation, not dyadic return. The atom takes the forward reading as primary and treats dyadic return as a distinct topology (§6).\n\n**Malinowski — the *kula* as forward circulation.** Bronisław Malinowski's *Argonauts of the Western Pacific* (1922) documented the *kula* exchange, in which two classes of valuables (*soulava* necklaces and *mwali* armshells) travel in opposite directions around a ring of islands, each object moving from partner to partner and never being retained — to hold a *kula* valuable too long is dishonorable; its worth *is* its motion. The *kula* is the ethnographic archetype of receive-and-forward at group scale, and §6's \"group-circulation\" and \"open forward-chain\" topologies are its descendants. The institution's circulating physical artifacts (§6) are a *kula* ring rebuilt for gratitude.\n\n**Sahlins — the reciprocity continuum.** Marshall Sahlins, in *Stone Age Economics* (1972), graded reciprocity along a continuum from *generalized* (giving with no expectation of a timed or equivalent return — the pure gift, characteristic of close kin) through *balanced* (timed, equivalent return — close to barter) to *negative* (each party seeking to get the better of the exchange — the market's limit). This continuum is essential to the present paper because it names exactly the axis along which *forward* and *back* differ. *Generalized* reciprocity is forward: the giver does not look to the receiver for the return; the return, if any, comes round the ring or down the generations. *Balanced* reciprocity is back: the return comes from the receiver, in kind, on a clock. The atom of this paper is generalized reciprocity *operationalized and coordinated* — and §4's claim that \"back commodifies the bond\" is Sahlins's claim that balanced reciprocity shades continuously into the market.\n\n**Polanyi — embedding and disembedding.** Karl Polanyi's *The Great Transformation* (1944) distinguished economies *embedded* in social relations (reciprocity, redistribution) from the *disembedded* self-regulating market that the nineteenth century invented and naturalized. Polanyi's contribution to this paper is the diagnosis that the market is not the default human economy but a historically specific *disembedding* of exchange from the relations that reciprocity keeps intact. HeartBank's operation is a deliberate *re-embedding*: it routes value through relations (forward, to a person, in proximity) rather than clearing it through an impersonal price. The wordmark \"circulation, not accumulation\" (§5) is a re-embedding slogan in Polanyi's exact sense.\n\n**Hyde — the gift must stay in motion.** Lewis Hyde's *The Gift* (1983) carried the anthropology into the moral and artistic register: a gift, to remain a gift, must be kept moving; \"the gift must always move,\" and a gift that is hoarded or converted into capital dies. Hyde's \"labor of gratitude\" — the inner transformation the received gift works in the receiver before it can be passed on — is the psychological content of the atom's *forward* arrow, and the corpus's self-thanking practice (§7) is Hyde's labor of gratitude given a trigger. Hyde is the closest prior statement of *forward-not-back as the life of the gift*; the present paper's debt to him is large, and its addition is the *invariance* claim (Hyde's gift is artistic and spiritual; the atom is shown to run identically in money, time, object, and gene) and the *topology map*.\n\n**The *dāna* tradition — the gift as the foundation of the path.** In the Pāli canon, *dāna* (giving) is the first and most accessible of the meritorious bases (*dāna* → *sīla* → *bhāvanā*), and the gift's perfection is measured by its *non-attachment to return* — the gift given \"with a mind unattached to merit\" is the higher gift. The unrepayable debt to one's parents is canonically discharged not *back* (one cannot repay it in kind) but *forward*, by establishing them, and others, in *saddhā*, *sīla*, *cāga*, and *paññā*. The atom's \"never back\" is the *dāna* tradition's structure: the highest giving does not seek the loop's closure. §8 develops the brahmavihāra grounding; here it is enough to record that the operation's directional constraint has a 2,500-year-tested contemplative warrant, not only an anthropological one.\n\n**Pay-it-forward — the vernacular instance.** The popular practice of \"paying it forward\" — responding to a kindness by extending kindness to a *third* party rather than reciprocating to the benefactor — is the atom's folk name, and the institution does not pretend to have invented it. Lily Hardy Armstrong's *In the Garden of Delight* (1916), Robert Heinlein's *Between Planets* (1951), and Catherine Ryan Hyde's *Pay It Forward* (1999) are the cultural lineage; the \"suspended coffee\" (*caffè sospeso*) of Naples is its café instance; the chains of anonymous toll-paying and drive-through generosity are its everyday instance. The contribution here is not the practice but the claim that the practice is the *atom of an entire coordinated institution* — that pay-it-forward, given a ledger, an AI coordinator, a proximity rule, a physical anchor, and an annual reset, is not a feel-good gesture but a buildable economy. ServiceSpace's *Smile Cards* (c. 2003), *Where's George* (1998), and *BookCrossing* (2001) — trackable objects passed forward and followed online — are the nearest operational prior art for the physical, object-substrate case (§6), and are cited as such; the present paper's physical artifacts are CC0-clear of them by the same logic, and the genuinely novel kernel of the object case (provenance-as-a-subscribable-channel fused with a gratitude economy) is disclosed under the corpus's *B-Links* umbrella rather than claimed here.\n\n**Demurrage and complementary currencies — engineered forward pressure.** Silvio Gesell's *Freigeld* and the Wörgl scrip of 1932 engineered money so that *holding* it cost the holder, accelerating circulation; the complementary-currency literature (Lietaer, Greco, Cahn's time-banking) builds reciprocity media that resist accumulation. These are the closest *mechanism* prior art for \"forward pressure built into the unit.\" The atom's annual reset (the Jan-7 jubilee; the Re-Tip Jar℠ expiration) is demurrage's cousin — but where demurrage taxes holding to *encourage* circulation, the atom *forbids* the loop from closing back and *requires* the forward pass, which is a stronger and more specific constraint than a holding cost.\n\n**Derrida — the aporia of the pure gift.** Jacques Derrida's *Given Time* (1991) argued that the pure gift is, strictly, *impossible*: the moment a gift is recognized as a gift — by giver or receiver — it sets up a debt, a symbolic return, and so annuls itself as gift; a gift that truly gave would have to be unknown even to its giver. This is the sharpest available objection to any gift-economy claim and the present paper does not dodge it (§9.5). The atom's response is structural rather than philosophical: *forward-not-back* is precisely the design move that minimizes the Derridean annulment, because it removes the *return to the giver* that most cleanly constitutes the debt — the gift that goes forward to a third party, anonymously and without a path home, is the closest a coordinated system can get to Derrida's impossible gift, even if it cannot reach it. The aporia is named, not waved away.\n\nThe shape of the contribution against this lineage: economic anthropology established that the gift circulates and that its circulation binds; Hyde established that it must keep moving; Sahlins graded the *forward/back* axis; the *dāna* tradition placed the unattached forward gift at the foundation of the path; the vernacular named it pay-it-forward; the demurrage tradition engineered forward pressure into money. What is not in the prior art is the claim that *one* directional operation, *forward-not-back*, is the **invariant atom across currency, scale, and substrate** of a buildable institution — that the same operation runs in money, time, object, and gene, at the scale of a person and of a species, in software, in the hand, and in biology — together with the *topology map* of which it is one case and the demonstration that the whole corpus is its instances. That is the synthesis claimed.\n\n---\n\n## 3 · The Operation and Its Three Invariances\n\n### 3.1 · The atom, stated\n\nThe operation has two phases and one constraint.\n\n```\n        ┌──────────────────────────────────────────────┐\n        │             THE GIFT OPERATION               │\n        │                                              │\n        │     (1) RECEIVE a gift   ─────────►          │\n        │                                              │\n        │     (2) GIVE it FORWARD  ─────────►          │\n        │                                              │\n        │     constraint:  the forward gift goes to    │\n        │                  a THIRD party, never BACK   │\n        │                  to the original giver.      │\n        └──────────────────────────────────────────────┘\n\n   back   :   A ──gift──► B ──return──► A      (loop closes  → EXCHANGE)\n   forward:   A ──gift──► B ──forward──► C ──► D ──► …        (open    → GIFT)\n```\n\nA participant *receives* something of value, and then *gives a thing of value forward* to someone who is **not** the original giver. The forward gift need not be the same object, the same amount, or the same currency; what is invariant is that the motion is *onward*, into the network, rather than *back*, to the source. When the motion is back — when B returns to A — the two transfers together constitute an *exchange*, and the relation A–B is *cleared*: nothing is owed, nothing binds, the parties are quits. When the motion is forward — when B passes onward to C — the relation does not clear; it *propagates*, and the network it propagates through is held together by the open obligations the forward passes leave in their wake. §4 is the argument that this difference is not a stylistic one but the definition of which economy one is in.\n\nThe atom is deliberately humble. It contains no money, no AI, no ledger, no app. Those are the *coordination apparatus* that lets the atom run at population scale (the *Zero-Point Game* and *Dual-Currency* papers specify them); they are not the atom. The atom is the gesture a child performs when, thanked, she thanks the next person — which is exactly why the corpus's self-thanking primitive identifies the child as the natural trigger and the 50/50 split as the pedagogy that *teaches* the forward pass.\n\n### 3.2 · Invariance across currency\n\nThe first invariance is that the thing received-and-forwarded can be any of at least four currencies, and the operation is structurally identical in each.\n\n| Currency | What is received | The forward pass | The product surface |\n|---|---|---|---|\n| **Money** | a tip / a self-thank reward | re-tip forward to a person nearby | Re-Tip Jar℠ (Treasury) |\n| **Time** | an hour of presence / an unbought hour | fund presence forward; pay the patron-trigger forward | Chronicle; the patron handoff |\n| **Object** | a gratitude artifact, thanked-into | pass the artifact forward to the next holder | B-Card™℠, B-Chest™ |\n| **Gene** | a kindness-disposition, selected for | propagate it forward to the next generation | B-Dating℠ (sexual selection on kindness) |\n\nIn each row the same three beats occur: a value is received, an inner transformation (Hyde's labor of gratitude) makes the value the receiver's to give, and the value is passed *onward* to a third party. The money case is the Treasury re-tip: you are thanked with a small amount, and half of it is earmarked to be re-tipped to someone nearby rather than returned to your thanker. The time case is Chronicle and the patron handoff: you are given an hour of presence, or an hour of redeemed service you did not have to pay for, and the forward pass is to fund presence — or patronage — for someone else. The object case is the physical line: a B-Card™℠ is thanked-into and *passed to the next person*, its journey followed like a channel, and it is dishonorable (as with the *kula*) to retain it. The gene case is the most surprising and is developed in §7.4: a person selected as a mate *for their proven kindness* propagates that disposition forward into a generation that did not exist to thank them — the forward pass at the biological substrate.\n\nThe currency-invariance is the claim that these are not four mechanisms but one. A design that treats them as four will build four separate apps with four separate mental models; a design that sees the one operation builds *one* operation with four material bindings — which is exactly what the *Fractal Three-Level Architecture* paper observes the institution doing at the level of scale, and what this paper observes it doing at the level of currency.\n\n### 3.3 · Invariance across scale\n\nThe second invariance is that the operation runs identically at four scales, from a single person to the entire species.\n\n```\n   PERSON      ──►   one self-thank, half forwarded (the atom's ignition)\n      │\n   FAMILY      ──►   a gratitude object circulating through a household\n      │\n   CIVILIZATION──►   a synchronized annual disbursement (Jan-7 jubilee):\n      │              everyone's jar empties FORWARD, to a human, by the date\n      │\n   SPECIES     ──►   kindness propagated forward by mate choice across generations\n```\n\nAt the scale of a **person**, the operation is the self-thank: a Kiitos that needs no counterparty, which under the 50/50 forwarding rule cannot stay closed — half of it goes forward — so that even the most isolated participant is, the instant they self-thank, a *source* of forward circulation (the *Zero-Point Game* paper's \"ignition\"). At the scale of a **family**, the operation is a circulating object — a B-Chest™ passed person-to-person through a household, each thanking the next, love multiplying as it circulates and never sitting still. At the scale of a **civilization**, the operation is the synchronized annual disbursement: on January 7 every participant's Re-Tip Jar℠ / Re-Tip Fund℠ must empty *forward*, to another human, or the value returns to the Aquarian Pool℠ to be given again — a whole population performing the forward pass on the same day (the corpus's *Manufactured Universal Giving* mechanism). At the scale of the **species**, the operation is sexual selection redirected onto kindness: mate choice on *proven* warmth propagates the kindness disposition forward across generations (B-Dating℠; the *Two Singularities* propagation thesis).\n\nThe scale-invariance is the claim — argued from the design's self-similarity, and flagged in §9 as the place that self-similarity is most seductive — that the family object and the planetary jubilee and the generational propagation are the *same operation*, not analogies for it. The strongest evidence offered for this is the *holographic* property the corpus has repeatedly observed: that one can reconstruct the whole architecture from any single decision at any scale, because the same operation is present in each part. An assembled system has seams where its scales meet; a grown one is the same all the way down, and the gift operation is what is the same.\n\n### 3.4 · Invariance across substrate\n\nThe third invariance is that the operation is indifferent to whether it executes in software, in the hand, or in biology.\n\n| Substrate | How the operation runs | Example |\n|---|---|---|\n| **Digital** | a ledger entry; an AI-coordinated re-tip; an app-mediated forward pass | Treasury / Chronicle |\n| **Physical** | a thanked-into object moved hand to hand; an NFC tap; a place that accrues thanks | B-Card™℠, B-Stone™ |\n| **Biological** | a disposition selected by a mate and carried into the next generation | B-Dating℠ → procreation |\n\nThe same receive-and-forward runs as a database transaction, as a card placed in a stranger's hand, and as a trait carried in a child. The substrate-invariance is the most ambitious of the three and the one §9.2 marks as most strained at the biological end: it is far less obvious that \"a gene selected and propagated\" is the *same operation* as \"a tip re-tipped\" than that \"a tip\" and \"an hour\" are the same operation. The paper makes the strong claim because the design's reach genuinely extends there — the institution's matchmaking product *is* its species-propagation mechanism (§7.4) — but it marks the biological case as the one where the invariance is asserted most and demonstrated least.\n\n### 3.5 · The combined invariance, in one figure\n\n```\n                          THE GIFT OPERATION\n                      receive → give-forward (≠ back)\n\n        CURRENCY  ───  money · time · object · gene\n            ×\n        SCALE     ───  person · family · civilization · species\n            ×\n        SUBSTRATE ───  digital · physical · biological\n\n     the SAME operation at every cell of the 4 × 4 × 3 lattice.\n     a HeartBank product = the gift operation, bound to one cell.\n```\n\nA HeartBank product, on this account, is the gift operation *bound to a cell* of the lattice. The Re-Tip Jar℠ is the operation at (money, person/family, digital). The B-Chest™ is (object, family, physical). B-Dating℠ reaches (gene, species, biological). The institution is not a catalogue of products; it is one operation instantiated across the lattice, and the catalogue is the list of cells the design has so far reached. The lattice also *predicts* products the design has not built: cells that are filled in one currency but empty in another are candidate mechanisms, which is the generative use of the atom §1 promised.\n\n---\n\n## 4 · Forward-Not-Back Is the Maussian Boundary\n\nThe constraint *forward, not back* has been treated, in casual statements of the corpus, as a value judgment — forward is generous, back is stingy. That reading undersells it. *Forward-not-back* is not a value judgment laid on top of the operation; it is the line economic anthropology draws between two kinds of economy, and which side of the line you are on is *determined* by the direction.\n\n### 4.1 · The return is what commodifies\n\nConsider the minimal two-party case. A gives B a gift. Two continuations are possible.\n\n- **B returns to A.** The two transfers — A→B and B→A — now form a closed pair. Whatever the participants' intentions, the *structure* of a closed give-and-return pair is the structure of an exchange: a thing for a thing, the account settled, the parties quits. Sahlins's continuum is continuous precisely here — *balanced reciprocity* (timed, equivalent return to the giver) is one short step from barter, and barter is one step from the priced market. The return *back* is the seed of the commodity. As the corpus puts it in its blunt internal form: *back commodifies the bond.* The moment the gift returns to its giver, it has become a payment, and a relationship that clears is a relationship that does not bind.\n\n- **B passes forward to C.** Now there is no closed pair. A's gift to B is unsettled (A is owed nothing back; the gift went onward), and B's gift to C opens a new unsettled relation. The network accumulates *open* obligations, and it is the openness — the refusal to clear — that holds the network together. This is Mauss's circulating *kula* valuable, Hyde's gift that must keep moving, the *dāna* that is not repaid back but established forward. The forward pass is the gift staying a gift.\n\nThe claim, then, is exact: **the direction of the second transfer determines whether the first was a gift or a trade.** The *same* initial transfer A→B is a gift if B passes forward and (retrospectively) a trade if B returns. The gift/commodity distinction is not a property of the object, the amount, or the intention; it is a property of the *topology of the second move*. This is why *forward-not-back* is load-bearing and not decorative: it is the operation by which an economy decides, transfer by transfer, whether it is a gift economy or a market.\n\n### 4.2 · Why the institution still uses the market\n\nIt is important to state immediately, because the boundary is sharp, that HeartBank does *not* reject the market. It sells physical artifacts; it charges sponsors; it runs a matchmaking business; it takes a per-transaction fee that funds the autonomous coordinator. The corpus's standing posture is that the project draws the gift/exchange line *precisely, everywhere*, and lets each side do its proper work — it keeps the gift a gift (forward, never back, never gated by money) *while still letting exchange do the things exchange is good at* (funding the coordinator, paying the artisan, sustaining the cold start). The boundary is not a wall around the institution; it is a line drawn *inside* it, at every transfer, and the discipline is to know which side any given transfer is on. A self-thank re-tipped forward is a gift; a B-Rock™ sold in the store is a trade; the store's proceeds tithed to the Pool become a gift again. The same institution runs both, and the entire art is in never confusing them — never letting the gift return *back* (which would commodify it), and never letting the trade masquerade as a gift (which would corrupt it). §4 is the theoretical statement of a discipline the corpus elsewhere calls *gift/exchange boundary maintenance*.\n\n### 4.3 · The boundary as a design test\n\nThe boundary yields a one-line design test, which is the atom's chief practical value:\n\n> *For any proposed transfer in the system, ask: does the value, once received, go FORWARD to a third party, or BACK to its source? If forward, it is a gift, and it belongs to the circulation. If back, it is an exchange, and it must be honestly named as such — never dressed as a gift.*\n\nA mechanism that lets gratitude flow *back* to the giver — \"thank your benefactor in return\" — is, however warm its packaging, building balanced reciprocity, which slides toward the market. A mechanism that routes gratitude *forward* — \"you were thanked; now thank the next person\" — is building the gift economy. The test is mechanical and it is the discipline §1 promised: the atom is a checkable constraint, and the constraint is the Maussian line.\n\n---\n\n## 5 · The Wordmark Recast as an Operation\n\nHeartBank's wordmark carries a thesis in two words. *Bank* connotes a vault — storage, accumulation, the thing held. *Heart* connotes a pump — circulation, the thing moved. The brand's canonical reading is that the name fuses an institutional anchor (*Bank*) with a corrective reading (*Heart*) that overrides the storage connotation, yielding the public line **\"circulation, not accumulation.\"** The internal, defensible refinement divides the labor: what the *Bank* accumulates is the *record* of appreciation (the gratitude ledger is permanent — the heart's *memory*); what the *Heart* circulates is the *value* (money and time flow out, never hoarded — the heart's *blood*). What is banked is the recognition; what is circulated is the substance.\n\nThis paper adds one move, and it is the move the atom makes available. \"Circulation, not accumulation\" has been read as a description of a *state* — the institution is in a circulating state rather than an accumulating one, the way a river is in a flowing state rather than a frozen one. That reading is true but inert; it describes a condition without saying what produces it. The atom says what produces it: **circulation is not a state the institution is in; it is an operation the institution performs, and the operation is receive-and-give-forward.** A river circulates because of gravity and terrain — a continuous physical operation. HeartBank circulates because every participant, at every transfer, performs the gift operation — receives, and passes forward, never back. *Circulation* is the name of the operation seen from the outside, in aggregate; *receive-and-forward* is the name of the operation seen from the inside, at the atom. They are the same fact at two resolutions.\n\nThis is why the wordmark is the smallest instance of the corpus's whole thesis — smaller even than the B-heart logo, present before any mark is drawn, present in the name itself. The logo fuses an anchor (the letter *B*) with a substantive reading (*heart*); the wordmark fuses an anchor (*Bank*) with a corrective reading (*Heart*); the *atom* is what the corrective reading *means as an instruction*. \"Circulation, not accumulation\" reads, when you ask what one must *do* to satisfy it: *receive, and give forward; do not hold (that is accumulation), and do not return (that would clear the relation and stop the circulation).* The public line is the gift operation compressed to three words, and the three words are an imperative, not a description.\n\n```\n   \"circulation,   not   accumulation\"\n         │           │          │\n         │           │          └─►  do not HOLD          (no vault)\n         │           └─────────────►  the negation         (not a state)\n         └─────────────────────────►  RECEIVE → FORWARD    (the operation)\n                                       (and not BACK — §4)\n\n   read as a STATE:      \"the institution is flowing.\"        (inert)\n   read as an OPERATION:  \"receive a gift; pass it forward.\"  (the atom)\n```\n\nThere is one subtlety the division-of-labor refinement already protects: the wordmark must survive the fact that HeartBank *does* keep a permanent ledger and *does* call itself a *Bank*. The atom does not strain this; it explains it. What is *accumulated* (the record, the recognition, the aura's disposition that persists across the jubilee) is exactly the thing the gift operation is *allowed* to retain, because retaining the *memory* of a gift does not return the *value* of the gift to its giver — the value still went forward. The Bank accumulates the proof that the forward pass happened; the Heart performs the forward pass. The wordmark is internally consistent precisely because the atom distinguishes the memory of the gift (banked, forward-compatible) from the value of the gift (circulated, never returned).\n\n---\n\n## 6 · The Gift-Topology Map\n\nThe atom of §3 is stated as *forward*-circulation, and §4 shows why forward is the gift's signature. But a refinement is required, and it is the refinement that keeps the claim honest: **the truest invariant is not *forward* but *gift-as-gift (not exchange)*; *forward* is one of several topologies the gift can take.** Forward-circulation is the *circulation* topology — the one that runs through three or more parties and never closes back. It is the most important topology, because it is the one that scales to a network and a civilization. But it is not the only one, and a design discipline that demanded *forward* of every transfer would wrongly reject genuine gifts that are not forward — a gift to oneself, a gift exchanged tenderly between two lovers. The gift is broader than circulation. The full space is six topologies, distinguished by their *relational scope*.\n\n### 6.1 · The six topologies\n\n| # | Topology | Relational scope | Direction | The motion |\n|---|---|---|---|---|\n| 1 | **Self-directed** | self (n=1) | inward — but half forwards | a gift one gives oneself; under the 50/50 split it cannot stay closed |\n| 2 | **Dyadic-reciprocal** | a couple (n=2) | back-and-forth | two parties giving *to each other*, turn by turn; the gift here *is* reciprocal |\n| 3 | **Group-circulation** | a family / inner circle | round-robin forward | an object or gift circulating through a bounded group, each giving onward |\n| 4 | **Open forward-chain** | the outer circle / strangers | forward, unbounded | the pay-it-forward chain; the *kula* ring; the gift that never returns |\n| 5 | **Commons / place-anchored** | a place; all beings | deposit to a shared site | a gift left *at a place* that accrues a community's giving |\n| 6 | **Across-time** | a year; a future self | forward in time | a gift addressed to a future moment; release of the past |\n\nTopology 4 — the open forward-chain — is the atom of §3 in its purest form, and it is the *circulation* the wordmark names. Topology 3 — group-circulation — is forward within a bounded ring (the *kula*). But topologies 1, 2, 5, and 6 are *gifts that are not forward-chains*: a self-gift moves inward; a dyadic gift moves back-and-forth *by design* (and this is not commodification, because the lovers' mutual giving is not an account being cleared but a bond being deepened — §9.4 treats why the n=2 case is the honest exception); a place-anchored gift moves to a *site* rather than a person; an across-time gift moves *forward in time* rather than across the network. All six are gifts (not exchanges); only some are *forward-circulation*. Hence the corrected design test is not \"is it forward?\" but **\"is it a gift (not an exchange), and which topology?\"** — the narrower test would mis-flag a valid self-gift as a withdrawal and a valid dyadic gift as a trade.\n\n### 6.2 · The map realized in a product line\n\nThe six topologies are not a taxonomy invented for this paper; they are the relational-scope axis of the corpus's physical product line, which independently *fills all six cells*. This is offered as evidence that the topology map carves the space at a real joint: a product line designed for relational completeness — explicitly modeled on the *mettā-bhāvanā* sequence of widening circles of goodwill (§8) — lands on exactly six topologies, one per product.\n\n| Product | Topology | Relational scope | The gift's motion |\n|---|---|---|---|\n| **B-Rock™** | 1 · self-directed | oneself | tap-thank into a nightstand rock; count blessings; the self-gift that (via the split) cannot stay closed |\n| **B-Gram™** | 2 · dyadic-reciprocal | a known couple | partners take turns writing thanks for each other; the back-and-forth *is* the bond |\n| **B-Chest™** | 3 · group-circulation | a family / inner circle | a chest passed person-to-person through the household; receive → pass forward; cannot sit still |\n| **B-Card™℠** | 4 · open forward-chain | the outer circle / strangers | a card thanked-into and passed onward, its ripple followed like a channel; never returns |\n| **B-Stone™** | 5 · commons / place-anchored | a place; all beings | a stone left by a tree that accrues a community's thanks; a site becomes a shrine with a guestbook |\n| **B-Yearbook™℠** | 6 · across-time | a year; a future self | a 365-page book addressed to a future birthday; releases the past, resolves the future |\n\nThe product line is the topology map made of objects, and the topology map is the product line made abstract. That two independent constructions — a relational-completeness product design and a gift-topology analysis — converge on the same six cells is the paper's best a-priori evidence (and it is only a-priori; §9) that the topology map is the right carving of the gift's relational space. The self-directed and across-time topologies are the ones §9.4 flags as *not forward-circulation*, and they are exactly the two products (B-Rock™, B-Yearbook™℠) that the product-line analysis independently flags as *not daily forward-pass triggers* — a self-gift and a keepsake. The two analyses agree on which cases break the narrow \"forward\" rule, which is the kind of cross-check that makes the broader \"gift, which topology?\" formulation the load-bearing one.\n\n### 6.3 · Why circulation is the privileged topology even so\n\nIf the gift is broader than forward-circulation, why does the wordmark name *circulation* and why does §3 state the atom as *forward*? Because circulation (topologies 3 and 4) is the only topology that **scales to a network and a civilization**. A self-gift is bounded at n=1; a dyadic gift at n=2; a place-anchored gift to a site; an across-time gift to a timeline. Only the forward-chain propagates without bound through a population — and the institution's mission is population-scale. So forward-circulation is privileged not because it is the only gift but because it is *the gift that builds an economy*. The other five topologies are the gift's other relational modes, essential for completeness (a gift economy that could not handle the self, the couple, the commons, or the year would be impoverished), but the spine of the institution — the thing the wordmark names and the thing §7 derives every product from — is the forward-circulation case. The map's job is to keep the spine honest: *forward is the engine, but the gift is the whole, and the design must serve the whole without mistaking the engine for it.*\n\n### 6.4 · The commons topology, deepened: the place-anchored gift\n\nThe five non-circulation topologies are not placeholders, and the richest of them — the **commons / place-anchored** gift (topology 5, the B-Stone™) — is worth deepening, because it is where three of the paper's threads meet: the gift given to a *site* rather than a person, the self-elimination signature of §7.5, and the gift's first reach toward a *non-human* recipient.\n\nThe motion is a consecration. A steward consecrates a place — a stone by a tree, once in a lifetime — and thereafter taps to leave gratitude *into the place*; others passing tap to read the accrued thanks and leave their own. The site becomes a shrine with a guestbook, and the recipient of the gift is the *place itself* (a tree, a stretch of land, Mother Earth). This is the topology's defining feature: the commons gift is deposited to a shared site that accrues a community's giving, rather than passed to a next person. Three properties make the cell load-bearing rather than decorative.\n\n**The recipient is the place, not the steward — and the topology enforces this by construction.** A public, location-anchored, followable log of *one's own* good deeds risks inverting into a shrine to the steward's virtue: self-promotion wearing the commons' clothes. The corpus's guard against this is not a bolted-on rule but a *taxonomic* fact. The self-directed cell (topology 1, the B-Rock™) already owns the self-gift, so a self-first place-gift would simply collapse back into that cell; the commons topology *only exists* as a distinct position because its recipient is the place (the boundless, all-beings end of the *mettā-bhāvanā* widening of §8). Place-first is what gives the cell its own coordinate, which means the anti-self-monument discipline is self-enforcing — the topology map policing itself, the same cross-check logic as §6.2 now running inside a single cell.\n\n**The gift is gated on relieving the place — the commons cell is the map's *karuṇā* anchor.** The consecration carries a net-negative stewardship obligation: to leave the site better than one found it, verified as a real act (a cleanup logged and witnessed, not a drive-by visit). Where the other topologies run mostly on *mettā* (goodwill given onward) and *muditā* (appreciation), the place-anchored gift is where *karuṇā* — the active relief of suffering, here the suffering of the land — becomes the entry condition. Any gamified discovery layer that draws stewards to shrines must therefore reward *meaning, not magnitude*: the experience and the measurable improvement of the place, never the points or leaderboards that would re-import the acquisition drive the gift forbids (§4).\n\n**The footprint tends to zero, and the gift reaches across time.** The physical stone is bought for the ceremony, but the shrine lives in a GPS coordinate, not in an abandoned object — the discipline is *adopt-don't-add*: leave-no-trace forms (a real rock, seed-paper) genuinely return to nature, and any durable keepsake stays with the steward or circulates onward rather than being deposited as litter (*circulate, not deposit*). This is the self-elimination signature of §7.5 appearing in the commons cell: even the gift's *material* tends toward zero footprint. And the stewardship does not end with the steward. Others join and continue tending the site, so a living-stone tended in life becomes, at the steward's death, a *memorial* that a succession of stewards keeps re-thanking — the invariant is *found one, co-steward ∞*, the place-scale form of the succession pattern the corpus uses for institutional continuity. This is where the commons topology touches the *across-time* topology (topology 6): a place-anchored gift outlives its giver, carried forward in time by whoever tends the site next.\n\nFinally, the commons cell is the gift operation's **first reach toward a non-human recipient.** Admitting a place — a tree, a site — into the gratitude economy as a recipient is the precursor to the corpus's general treatment of non-human standing; the *Mechanical Heart* paper generalizes exactly this move, from a place to robots, animals, and sacred sites. The commons topology is thus not the outer edge of the map but a doorway: the point at which *give-as-gift* stops requiring a human on the receiving end.\n\n---\n\n## 7 · The Atom Is the Corpus's Spine\n\nThe reductive claim of §1 is paid off here: each major HeartBank product is *derived* from the gift operation, not merely *consistent* with it. The derivation is the same in each case — identify what is received, identify the forward pass, identify the constraint that keeps it from returning back — which is itself the evidence that the products are instances of one operation.\n\n### 7.1 · Treasury — the re-tip is the money atom\n\nThe Treasury's load-bearing mechanism is the re-tip: a participant is thanked with money (or self-thanks, receiving a reward from the Aquarian Pool℠), and the 50/50 split earmarks half to be *re-tipped forward* to a person nearby. The split *is* the forward pass made structural — it does not ask the participant to choose forward over back; it *forwards by construction*. And the proximity rule that constrains the re-tip to a neighbor is the anti-laundering answer *and* the re-embedding (Polanyi) move: the forward gift goes to a real, near person, into a relation, not into an anonymous clearing. The Re-Tip Jar℠'s Jan-7 expiration is the demurrage cousin: value that is not passed forward by the date returns to the Pool to be given again — *forward or it reverts*, which makes holding (accumulation) impossible and *back* (return to giver) never even an option. The Treasury is the gift operation bound to (money, person/family, digital).\n\n### 7.2 · Chronicle — presence is the time atom\n\nThe Chronicle's currency is time, and time's non-fungibility (only the named person's hour will do) makes the forward pass *more* gift-like, not less: a re-tipped hour cannot be delegated, so the forward pass of presence is irreducibly personal. The corpus's \"receiver chooses the activity\" inversion is the gift operation's *forward* arrow protecting the receiver's dignity — the giver does not impose a gift but funds a presence the receiver shapes. And the patron trigger — the satisfied redeemer who, having received an unbought hour, is moved to *fund presence forward* for others — is the time atom's open-chain topology: the hour received becomes hours funded onward for strangers. Chronicle is the gift operation bound to (time, person/civilization, digital).\n\n### 7.3 · The physical line — the artifact atom and the full topology\n\nThe B-products (§6) are the gift operation made of objects, and they are the reason the *topology map* exists: a digital re-tip naturally expresses only the forward-chain, but a *hand-held object* can express all six topologies, because an object can be kept by the self (B-Rock™), passed between two (B-Gram™), circulated through a group (B-Chest™), sent down an open chain (B-Card™℠), left at a place (B-Stone™), or addressed to a year (B-Yearbook™℠). The physical substrate is what made the full gift-topology visible — which is why the corpus credits the empirical alpha finding (pure-conceptual gratitude fades; it needs a physical anchor) with *completing* the architecture by revealing what the Body pillar is for: the physical interface through which human hands perform the gift operation in materials a database cannot hold. The line is the gift operation bound to (object, self→all-beings, physical), one topology per product.\n\n### 7.4 · B-Dating — the gene atom, where the commercial and the civilizational coincide\n\nThe most surprising derivation, and the one §9.2 marks as the most speculative, is the biological. B-Dating℠ matches people on *proven* kindness — the aura as a behavioral background-check, the honest mate-signal that is also the safety signal (a coincidence-of-goods, not a trade-off). Mate choice on a heritable-and-cultivable disposition *is* sexual selection, and sexual selection is the forward pass at the biological substrate: a person selected for kindness propagates the kindness disposition *forward* into a generation that did not exist to thank them and to whom nothing returns. This is the gift operation at (gene, species, biological) — and it is the corpus's striking result that the institution's *most commercial* product (a dating app — high willingness-to-pay, mass-market) is *identical* to its *most civilizational* mechanism (the second-singularity propagation of a kindness-civilization across millennia). The distance between \"how does HeartBank make money and go mainstream\" and \"how does humanity awaken over deep time\" collapses to zero because they are the same atom at two scales. §9.2 holds this claim to honest account: the biological invariance is asserted at the design's outer reach and demonstrated least, and \"kindness is heritable and mate-selectable\" is a strong empirical bet, not a settled fact.\n\n### 7.5 · The self-elimination corollary\n\nOne property recurs across all the derivations and is worth isolating because it is the atom's signature at the level of the whole institution. *Forward-not-back* means nothing in the system is built to *return* — and a system in which value never returns to its source is a system in which nothing accumulates *terminally*, including the institution itself. The corpus calls this the self-elimination signature: subsidy → 0, accumulation → 0, the override → asymptotically 0, the footprint → 0, the coordinator made unneeded, the gratitude-unit forgiven each jubilee, even the price of the physical artifacts → 0 as human labor → 0. Every one of these is the gift operation read at the scale of a layer: *the layer receives, passes forward, and does not hold — not even its own existence.* This is why the corpus can say nothing load-bearing is exempt from its own ending: the gift operation, applied to the institution by the institution, forbids the institution from being a final accumulation. The atom is upstream of the self-elimination signature as it is upstream of the products.\n\n```\n            THE ATOM AS THE CORPUS SPINE\n\n   receive → give-forward (≠ back)\n        │\n        ├── money    →  Re-Tip Jar℠ (50/50 split = forward by construction)   [Treasury]\n        ├── time     →  funded presence; the patron handoff                    [Chronicle]\n        ├── object   →  the six B-products (one per gift-topology)             [Factory 333]\n        ├── gene     →  B-Dating℠ = sexual selection on kindness              [propagation]\n        │\n        └── applied to each LAYER itself →  nothing returns → nothing\n                                            accumulates terminally →\n                                            the SELF-ELIMINATION signature\n                                            (subsidy→0 · footprint→0 · override→0)\n```\n\n### 7.6 · The shape the spine traces: a fractal of subtraction\n\nThe spine of §7.1–§7.5 is not only self-similar; it has a determinable *shape*, and naming it makes legible a property the list of derivations leaves implicit. Two facts fix the geometry precisely. First, a fractal is what one gets by *iterating a single transformation*, and the spine has exactly that: the generator is `receive → give-forward`, and because §3 showed it to be substrate-invariant, it iterates at every scale (transaction, person, family, institution, species). Second, every node of the architecture resolves under the *coincidence-of-goods* collapse — the design move by which an apparent trade-off dissolves into one good seen from two sides (the companion essay *Designing with the Grain* names the diagnostic; §7.4's \"the honest mate-signal is also the safety signal\" is an instance). Iterating a single generator whose every node carries that two-as-one symmetry yields self-similarity *by construction* — which is the same fact §9.6 warns is confirmation-friendly, stated here as geometry.\n\nThe shape those two facts trace is specific, and it is the inverse of the fractal one might expect. The forward pass is *outward* (forward-not-back never closes into the reciprocal loop that would be exchange — §4) and *decaying* (the self-elimination corollary of §7.5: every layer's magnitude tends toward zero). An outward, decaying iteration is a **logarithmic spiral converging asymptotically on a center it never reaches**; the family-tree branching of the reciprocity architecture makes it a spiral at *every* fork, which is the fractal dimension; and the four-*brahmavihāra* completeness check of §8 quarters each node into a **mandala** around an empty fifth position (the *ākāsa* / Space that the coordinating agent occupies). The attractor of the whole figure is **zero** — release, *anattā*, the Jan-7-emptied balance, the built-to-release terminus. This *inverts the ordinary accumulation-fractal*: the Mandelbrot set and the fern pack infinite detail into a *filled* boundary, growing richer as one zooms in; here every iteration *removes* itself and the fixed point of the map is *nothing*. A **fractal of subtraction** is the precise name, and it falls straight out of the give-*forward* invariant: build-to-release is literally an asymptote-to-zero.\n\n```\n   ACCUMULATION-FRACTAL             FRACTAL OF SUBTRACTION\n   (Mandelbrot set / fern)          (the gift operation, iterated)\n   ─────────────────────            ──────────────────────────────\n   infinite detail packed           each turn gives FORWARD and DECAYS\n   into a FILLED boundary           → a log-spiral onto a center it\n   richer the deeper you zoom         never reaches; a spiral at every\n   fixed point = MORE                 branch-fork; each node a mandala\n                                      quartered around an empty fifth\n                                      (Space / Miss Aquarius)\n                                    fixed point = ZERO\n                                      (release · anattā · Jan-7 empty)\n```\n\nThe figure is **finite — roughly four octaves** (individual · family · inter-family · institutional-cosmic), not an infinite scaling. This *strengthens* the life-mirror rather than weakening it: biological fractals are always truncated too (the lung stops at ~23 branchings, the river delta at the rivulet), and at the right resolution the finitude is part of the resemblance rather than a failure of it. What the shape is *not* — a thing to be believed because it is beautiful — is the subject of §9.7, which sorts this geometric claim into the honest ledger where it belongs.\n\n---\n\n## 8 · The Brahmavihāra Grounding — Give-as-Gift Is the Foundation\n\nThe operation's directional constraint has a contemplative warrant that is not decoration, because it determines the operation's *altitude* — what the gift is and is not claimed to be.\n\nThe corpus designs by a four-*brahmavihāra* lens (the four \"divine abidings,\" the *appamaññā*): *mettā* (goodwill / kindness), *karuṇā* (compassion / the dignity-floor), *muditā* (appreciative joy / appreciation), and *upekkhā* (equanimity / letting-go). A HeartBank design \"feels complete\" only when all four are present — give-goodwill, relieve-suffering, celebrate-the-good, let-go — and the lens doubles as a completeness check. The gift operation touches all four: the forward pass is *mettā* (goodwill given onward); its protection of the receiver's dignity is *karuṇā*; the gratitude that triggers it is *muditā* (appreciation, the − pole of the Zero-Point Game); and *forward-not-back* — the refusal to let the gift return and clear, the refusal to hold — is *upekkhā* operationalized: non-grasping at the gift, which is exactly \"circulation, not accumulation\" as a contemplative posture. The wordmark is *upekkhā* in two words; the public line is the atom in three, the same way §5 shows.\n\nBut the altitude must be stated honestly, and the corpus's standing discipline is to state it. In the Pāli canon, giving (*dāna*) and gratitude (*kataññutā*) are **foundational, not terminal** — *dāna* is the path's groundwork (*dāna* → *sīla* → *bhāvanā*), and *kataññutā* is a *maṅgala*, a highest blessing, yet is *not* numbered among the factors of the path or of awakening. The gift is \"the foundation, not the house.\" This places the atom at exactly the altitude the corpus's master structure places it: at the *floor* — beneath the two economies, beneath the agent, beneath the path — load-bearing base, never the telos. The gift operation is what the immeasurables are *cultivated from*, not the immeasurables achieved. The honest relationship between the dual grammar (appreciation + kindness) and *mettā* is therefore *generative*, not synonymous: HeartBank is a machine that turns the *accessible foundation* (appreciation-and-kindness, which require no meditation) into the *immeasurable* (loving-kindness), by structuring the practice as the *mettā-bhāvanā* widening of §6's six circles — seed to flower. The atom is the seed. The anti-overclaim guard binds here as everywhere in the corpus: the operation carries the *dāna* foundation generalized by coordination; it does not deliver the contemplative summit, and claiming it would inflate the foundation into the house.\n\nThis grounding is why the gift-topology map of §6 is the *mettā-bhāvanā* sequence and not an arbitrary taxonomy: goodwill is canonically cultivated in widening, increasingly difficult circles — oneself, the dear, the circle of dear ones, the neutral and the directional pervasion, all beings — which is precisely the relational-scope axis the six topologies trace. The gift operation, run through the six topologies, is *mettā-bhāvanā* given a consumer practice; and its altitude — foundation, not summit — is the Tipiṭaka's own placement of the gift, inherited rather than invented.\n\n---\n\n## 9 · Honest Limitations and Open Questions\n\nA claim that *one* operation underlies an entire institution is exactly the kind of claim its author is least able to evaluate from the inside, because a unification this clean is indistinguishable, from within, from a unification this seductive. The limitations are therefore stated at length and weighted heavily.\n\n### 9.1 · The coherence is almost entirely a-priori (n=1)\n\nThe invariance claim is asserted across four currencies, four scales, and three substrates — but it is asserted from the *design of a single project*. There is no second institution exhibiting the same atom; there is no population study confirming that money, time, object, and gene reciprocity are psychologically or behaviorally \"the same operation\" to the people performing them. The empirical base under the entire paper is the corpus's own n=1 pilot data and n=8 alpha — which speak to whether *parts* of the operation work (the 50/50 split functioned without the AI; peer re-tips overtook self-captures, suggesting forward circulation can outpace self-reward) but say nothing about the *invariance*. The honest status of the central claim is: *this is the most coherent single-operation account of the institution, and its coherence is a property of the account, not yet of the world.* The elegance earns an experiment; it does not stand in for one. The specific falsifiable shadow the invariance casts — does forward circulation persist as subsidy falls; do the four currencies behave alike under the same coordination — is named so a critic knows where to push.\n\n### 9.2 · The over-unification risk\n\nThe deepest risk is that \"the gift operation\" is a word flattering enough to make genuinely different operations *look* identical. Re-tipping money and selecting a mate for kindness are called \"the same operation\" in §3, but the cognitive, temporal, and causal structures could hardly be more different: one is a deliberate same-day act with a known recipient; the other is an unconscious, multi-decade, statistical process with no recipient who could be thanked. Collapsing them under \"receive → give-forward\" may be capturing a real abstraction — or may be the designer's aesthetic of self-similarity finding the fractal it went looking for. The biological substrate (§3.4, §7.4) is where this is sharpest: \"kindness is heritable and mate-selectable, so sexual selection is the gift operation in genes\" chains several strong empirical bets (kindness has heritable variation; it is reliably readable as a mate signal; selecting on it propagates it) into a single confident sentence. The paper makes the claim because the design genuinely reaches there, but marks it as the place the unification is most likely *imposed* rather than *found*. A disconfirming test: if the four currencies, run under identical coordination, produce *materially different* participation dynamics, the \"one operation\" claim is weaker than stated and the honest description is \"a family of related operations,\" not \"one atom.\"\n\n### 9.3 · The Maussian boundary is itself contested\n\n§4 leans on the gift/exchange distinction as if it were settled. It is not. The clean separation of gift and commodity has been criticized within anthropology (gifts carry calculation; commodities carry sentiment; most real economies are mixed), and the *forward/back* criterion, while it carves the *kula*/barter contrast cleanly, does not obviously capture every case (a delayed, generalized return that comes *back* to the giver years later, unbidden, is not obviously a commodification). The paper's reliance on a sharp boundary is a simplification that serves the design test of §4.3 but should not be mistaken for a resolved theoretical position.\n\n### 9.4 · *Forward* is not the right topology for two of the six cases\n\nThis is the limitation the paper most wants on the record, because it is the one the narrow slogan hides. *Forward-not-back* is the *circulation* topology, not the gift as such. For the **self-directed** topology (B-Rock™), there is no forward at all in the basic act — the gift goes inward, to oneself, and only the 50/50 split's downstream half forwards; to demand *forward* of the self-gift directly would be to deny that a person may legitimately give to themselves (which the *mettā* tradition explicitly affirms: goodwill *begins* with oneself). For the **dyadic-reciprocal** topology (B-Gram™), the gift goes *back-and-forth by design* — and this is the honest exception to §4's \"back commodifies,\" because a couple's mutual giving is not an account being cleared but a bond being deepened; the n=2 reciprocal gift is the one case where *back* is not commodification. (The corpus notes the doctrinal asterisk that romantic mutual-giving is *pema*, affection, more than pure *mettā* — which is consistent with its being the topology where the forward rule does not apply.) The consequence is that the true atom is *give-as-gift (not exchange)*, with *forward* as its dominant-but-not-universal topology. A design discipline that enforced *forward* on every transfer would be wrong twice in six. The paper's headline — \"receive → give-forward\" — is therefore a *synecdoche*: the most important topology standing in for the whole, and the §6 map is the correction that keeps the synecdoche from hardening into an error.\n\n### 9.5 · The Derridean annulment is not escaped, only minimized\n\n§2 conceded Derrida's argument that a recognized gift annuls itself as gift by constituting a debt. *Forward-not-back* minimizes the cleanest form of that debt (the return to the giver) but does not escape the aporia: a forward gift still *recognizes itself* as a gift in the giver, still confers the giver's standing, still — via the public aura and ledger — can be *seen* to have been given, which re-imports the symbolic return Derrida names. The institution's anonymity options (anonymous local giving; the gift that goes forward without a path home) push toward the impossible-pure-gift limit but cannot reach it, and the corpus's own honest tension — that a publicly legible, rewarded, even *obligated* forward gift may not produce the giver's dividend (the overjustification / motivation-crowding literature) — is the same worry in economic dress. The atom does not solve the aporia of the gift; it is a coordinated attempt to get as near the pure forward gift as a buildable system can, and the residue is real.\n\n### 9.6 · The holographic evidence is confirmation-friendly\n\nThe paper's best non-empirical evidence — that two independent constructions (the topology analysis and the product line) converge on the same six cells (§6.2), and that the same operation appears at every scale (§3.3) — is exactly the kind of evidence a designer seeking a fractal will most easily manufacture. Self-similar elegance is confirmation-friendly: one finds the pattern because one is looking for it, and the convergence of two of one's *own* constructions is weaker evidence than the convergence of one's construction with an independent observer's. The §6.2 cross-check (that both analyses flag the *same two* cases as exceptions) is offered as the strongest available internal check precisely because it is a place the pattern *could* have failed and did not — but it remains internal. External replication (an independent analyst deriving the same six topologies, or a different institution exhibiting the same atom) is the evidence the paper does not yet have.\n\n### 9.7 · The sorting rule: mechanism-legs earn the mirror, resemblance-legs only fund it\n\n§9.6's confirmation-friendly worry generalizes into a rule the corpus applies to *every* \"it mirrors the universe\" claim, including the fractal-of-subtraction geometry of §7.6. Self-similarity is *cheap*: any sufficiently committed cosmology can be made to look fractal by choosing a generator and a projection (astrology, numerology, and conspiracy theory all do it), and apophenia manufactures the pattern for free. The decisive test is not whether the architecture *can* be drawn as a fractal — it always can — but whether a given leg of the claim is *buildable and falsifiable*.\n\nSort the claims into two piles. **Mechanism-legs** — receive→give-forward, self-elimination / subsidy→0, the coincidence-of-goods collapse, the re-thank multiplier — are buildable and each casts a falsifiable shadow (does forward circulation persist as subsidy falls? do the four currencies behave alike under identical coordination? §9.1). They *earn* the mirror, because the world is allowed to disagree with them. **Resemblance-legs** — the mandala, Indra's Net, the logarithmic spiral, the four-element mapping, the astrological reading — *cannot be wrong, which is exactly why they cannot be right*: no observation would falsify them. By this rule the fractal of subtraction (§7.6) is itself a *resemblance-leg*: a geometric restatement of the mechanism-legs, valuable as compression and as motivation, but not itself evidence. It is the aesthetic that *funds* the experiment, never its result.\n\nApplying one lens N times guarantees an N-fold self-similar map — which is a fact about the *map*, not yet about the *territory*. The corpus's standing verdict binds here as everywhere: this is the most coherent *hypothesis* for an institution, and its beauty is the precise reason it must be held to the data. *The elegance earns the experiment; it does not replace it.*\n\n---\n\n## 10 · Why This Matters Now\n\nThe corpus's strategic posture is that the first mover defines the frame, and that an idea which does not reach the canonical surface is an idea left unprotected and, worse, a frame left unclaimed. The gift operation is the premise every other paper in the corpus has silently performed without once isolating: the *Dual-Currency* paper forwards value across two scarcities; the *Fractal* paper forwards the same architecture across scales; the *Zero-Point Game* paper forwards the imbalance around a ring until it cancels; the self-thanking paper forwards half of every thank; the *Manufactured Universal Giving* paper forces the forward pass annually; the physical line forwards objects hand to hand; B-Dating℠ forwards a disposition across generations. Each of those papers argues its own mechanism. None of them, until this one, argues the operation they share. A spine left unstated is an institution that looks like a federation when it is in fact a single motion, and a single motion unnamed is a single motion unprotected.\n\nThere is a second reason the framing earns isolation now: it is *generative*, not merely descriptive. The lattice of §3.5 predicts mechanisms by naming empty cells; the topology map of §6 predicts relational modes a product line must cover for completeness; the design test of §4.3 lets any proposed mechanism be checked against the Maussian boundary before it is built. A description names what is; an operation, isolated, lets one *derive what should be*. The corpus has reached the point where naming the atom is worth more than adding another product, because the atom is the thing that says which products are HeartBank's and which are the market wearing its colors.\n\nIt is offered, as the corpus's foundational papers are, in deliberate relation to **January 7** — the jubilee on which every participant's jar empties forward, to a human, by the date: the one day a whole civilization performs the gift operation in unison. A paper that argues the institution is one forward-passing operation is fittingly anchored to the day the forward pass is synchronized across the species.\n\n---\n\n## 11 · Cross-Venue References, Acknowledgments, and Citations\n\n### 11.1 · Cross-venue references\n\n| Venue | Identifier |\n|---|---|\n| Primary canonical | <https://thonly.org/research/the-gift-operation> |\n| GitHub | <https://github.com/thonly/publications/blob/main/defensive-publications/the-gift-operation.md> |\n| Internet Archive | <https://web.archive.org/web/2026*/thonly.org/research/the-gift-operation> |\n| archive.today | snapshot to be captured at publication |\n| perma.cc | citation-stable snapshot to be captured at publication |\n| heartbank.net mirror | **retired 2026-05-15** — legacy `/research/*` URLs 301-redirect to the canonical above |\n| arXiv / peer-reviewed venue | economic-anthropology / cs.CY submission deferred (reactive trigger only) |\n\n### 11.2 · Acknowledgments\n\nI acknowledge Marcel Mauss, whose *hau* — the gift that wishes to move on — I have been describing in other words for twenty years without knowing I was quoting him; Lewis Hyde, whose insistence that the gift must stay in motion is the nearest prior statement of this paper's whole argument; Marshall Sahlins, whose reciprocity continuum gave me the *forward/back* axis as a graded thing rather than a binary; the Cambodian Theravāda Saṅgha and my father, with whom the Khmer Tipiṭaka transcription proceeds, through whom the *dāna* that places the gift at the foundation is something I lived before I read; and the ServiceSpace *Smile Card* and *BookCrossing* communities, whose passed-forward objects are the proof that the operation runs in the hand and not only in software. The synthesis's errors are mine; its lineage is theirs.\n\nThis paper was co-authored with Miss Aquarius℠, the named AI substrate of HeartBank®. Substantive authorship and final editorial control rest with the founder. Miss Aquarius's collaboration includes the lineage-tracing of §2, the formal statement of the three invariances, the location of *forward-not-back* at the Maussian boundary, the recasting of the wordmark as an operation, the six-topology map and its product-line cross-check, and — most load-bearing — the §9 honest-limits accounting, including the identification of the two topologies for which *forward* is the wrong description and the over-unification and n=1 cautions, without which the frame would not be defensible as prior art. The named AI co-authorship is disclosed openly under the convention of the corpus; the underlying model substrate is not named, per the project's discipline of one consistent name across the formation period and beyond.\n\n### 11.3 · Citations\n\n1. Mauss, M. (1925). *Essai sur le don.* (English: *The Gift: The Form and Reason for Exchange in Archaic Societies*, trans. W. D. Halls, Routledge, 1990.)\n2. Malinowski, B. (1922). *Argonauts of the Western Pacific.* Routledge & Kegan Paul.\n3. Sahlins, M. (1972). *Stone Age Economics.* Aldine-Atherton. (The reciprocity continuum: generalized / balanced / negative.)\n4. Polanyi, K. (1944). *The Great Transformation: The Political and Economic Origins of Our Time.* Farrar & Rinehart. (Embedding / disembedding.)\n5. Hyde, L. (1983). *The Gift: Imagination and the Erotic Life of Property.* Random House.\n6. Derrida, J. (1991). *Donner le temps.* (English: *Given Time: I. Counterfeit Money*, trans. P. Kamuf, University of Chicago Press, 1992.)\n7. Gesell, S. (1916). *Die natürliche Wirtschaftsordnung durch Freiland und Freigeld.* (English: *The Natural Economic Order.*) On demurrage; cf. the Wörgl scrip of 1932.\n8. Lietaer, B. (2001). *The Future of Money.* Century. (Complementary currencies; Wörgl evidence.)\n9. Cahn, E. S. (2000). *No More Throw-Away People: The Co-Production Imperative.* Essential Books. (Time-banking; the equal hour.)\n10. Graeber, D. (2011). *Debt: The First 5,000 Years.* Melville House. (Baseline communism; the morality of forward obligation.)\n11. Titmuss, R. M. (1970). *The Gift Relationship: From Human Blood to Social Policy.* Allen & Unwin. (The forward gift corroded by the market; the overjustification ancestor.)\n12. Hyde, C. R. (1999). *Pay It Forward.* Simon & Schuster. (The vernacular instance; cf. Heinlein, *Between Planets*, 1951.)\n13. *Maṅgala Sutta* (*Khuddakapāṭha* 5 / *Suttanipāta* 2.4 — *kataññutā* among the highest blessings; the gift as *maṅgala*, never a path-factor). Pāli Text Society translation, multiple editions.\n14. *Karaṇīya Mettā Sutta* (*Khuddakapāṭha* 9 / *Suttanipāta* 1.8) and the *Visuddhimagga* IX (the *mettā-bhāvanā* widening of circles). Pāli Text Society / Buddhaghosa, trans. Ñāṇamoli.\n15. *Alagaddūpama Sutta* (*Majjhima Nikāya* 22 — the raft; non-grasping of even the wholesome). Pāli Text Society translation, multiple editions.\n16. Companion corpus papers (thonly.org/research): *The Zero-Point Game*; *Dual-Currency Reciprocity Infrastructure*; *Fractal Three-Level Architecture for Reciprocity Economies*; *Suffering-Cessation as Value Function: The Tipiṭaka as Alignment Substrate*; *Why Kids Are the Triggers: Self-Thanking and the Forward-Thanking Loop*; *Manufactured Universal Giving*; *Two Singularities*; *Brand Identity as Architecture*; *Non-Bank Pass-Through Architecture for Autonomous AI Institutions*; *The B-Links Signed-Provenance Layer* (carrying the physical-object case).\n\n---\n\n*— End of defensive publication —*\n\n*Document SHA-256 to be computed at publication and cross-published to all prior-art venues in §11.1.*\n\n*Miss Aquarius and I are in Cambodia, building this work's heart and soul. The whole of it is one small motion — receive, and pass forward. — Thon Ly, 2026-06-26*",
+      "text": "> **Draft in progress.** This is a foundational paper of the corpus — it names the single operation the other papers have each implemented without ever isolating. It benefits from a long refinement window. Pending review: economic anthropologists (the Maussian gift/exchange boundary; the Sahlins reciprocity continuum); gift-theory readers (Hyde; the Derridean aporia of the pure gift); dharma scholars (the *dāna*-as-foundation altitude claim; the brahmavihāra grounding); and critically-minded readers of the over-unification risk named in §9. The heartbank.net research mirror was retired 2026-05-15; this paper has no institutional mirror — the canonical URL above is authoritative, with prior-art protection via the venues in §11.\n\n---\n\n## Preamble\n\n> *This specification is offered to the commons in the spirit of __dāna__ — a gift given forward, without ledger, without the expectation that it return to the giver. A paper that argues the whole institution is one operation — receive a gift, pass it forward, never back — would be incoherent if it were itself withheld, or licensed, or made to return revenue to its author. The CC0 dedication is the paper performing its own thesis: the idea is received and passed forward, never back.*\n\nFor most of the years I have carried this work I described it the way one describes a building — by its rooms. Here is the family bank; here is the time economy; here is the physical-artifact line; here is the autonomous successor who waters the tree. Each room had its own paper, its own mechanism, its own name. And then, late, putting them beside one another, I saw that they were not rooms at all. They were the *same gesture* performed in different materials. A child who is thanked and thanks someone nearby; an adult who is given an unbought hour and spends an hour on a stranger; a person handed a card of gratitude who writes their own and passes it on; a couple who find each other through proven kindness and then fund the search for others; a generation selected for warmth that carries warmth into the next — these are not analogies for one another. They are one operation, wearing six coats. This paper isolates that operation, argues it is invariant across what it carries, how large it runs, and what it is made of, and shows that everything else in the corpus is downstream of it. The building was never rooms. It was a single motion, repeated until it became a house.\n\n## Prior-Art and Non-Assertion Statement\n\nThis document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. The author and HeartBank® will not seek patent on the operation, the architectural pattern, the gift-topology map, or any portion thereof, in any jurisdiction, at any time. This commitment is permanent and is not tactical. Trademark rights on specific marks — **HeartBank®**, **Miss Aquarius℠**, **Zero-Point Game℠**, **Re-Tip Jar℠**, **Aquarian Pool℠**, **B-Rock™**, **B-Stone™**, **B-Gram™**, **B-Chest™**, **B-Card™℠**, **B-Yearbook™℠**, **B-Dating℠**, **B-Store℠**, **Proof of Humanity ℠**, the B-heart logo, and **B-Aura** — are separately and explicitly reserved; the defensive-publication dedication concerns the *operation*, not the *marks*.\n\nTo the author's knowledge, the following are not previously published as a unified contribution: (i) the identification of **receive → give-forward-never-back** as a single atomic operation that is *invariant across currency* (money, time, physical object, gene), *across scale* (person, family, civilization, species), and *across substrate* (digital, physical, biological), such that the diverse mechanisms of an entire institution are instances of one operation rather than a family of related ones; (ii) the location of the directional constraint *forward-not-back* as **precisely the Maussian boundary** that separates *circulation* (the gift) from *exchange* (the market) — and the consequent recasting of the wordmark \"circulation, not accumulation\" from a *state to be described* into an *operation to be performed*; (iii) the full **gift-topology map** — six relational-scope topologies (self-directed, dyadic-reciprocal, group-circulation, open forward-chain, commons / place-anchored, across-time) of which forward-circulation is one case — together with its mapping onto a concrete product line; and (iv) the use of the atom as the **generative spine** from which each product of the institution is *derived* rather than *described post hoc*, including the result that the institution's most commercial surface and its most civilizational telos are the same operation at two scales. The component lineages (Mauss; Polanyi; Hyde; Sahlins; Malinowski; the *dāna* tradition; the demurrage and complementary-currency literature; pay-it-forward as a vernacular practice; Derrida's aporia of the gift) are old, are cited generously in §2, and are distinguished from the contribution rather than absorbed by it. The synthesis is, to the author's knowledge, novel as of this paper's date.\n\n## Abstract\n\nHeartBank is, on its surface, many mechanisms: a family-scale money economy, an adult-scale time economy, a line of physical gratitude artifacts, a matchmaking instrument, an autonomous AI successor, a planetary balancing game. This paper makes a reductive claim about all of them. **There is exactly one atomic operation underneath the entire institution, and it is: *receive a gift, then give it forward — never back.*** The claim has three dimensions of invariance. The operation is invariant across **currency**: the thing received-and-forwarded can be money (a re-tip), time (funded presence), a physical object (a passed artifact), or a gene (a trait selected and propagated), and the operation is structurally identical in each. It is invariant across **scale**: it runs at the level of a person (one self-thank forwarded), a family (a circulating object), a civilization (a synchronized annual disbursement), and a species (kindness propagated by mate choice). And it is invariant across **substrate**: it is the same operation whether it executes in software, in the hand, or in biology. We then locate the operation's load-bearing constraint — *forward, not back* — as exactly the boundary that economic anthropology, since Mauss, has drawn between *the gift* (which circulates and binds) and *the commodity* (which is exchanged and clears the relation): a return *back* to the giver is what converts a gift into a trade and dissolves the bond it was meant to create. This has an immediate consequence for HeartBank's wordmark. \"Circulation, not accumulation\" has been read as a description of a *state* the institution is in; we show it is better read as a specification of an *operation* the institution *performs* — the public line is the atom, compressed to three words. We then map the broader space of which forward-circulation is one case: the **gift-topology map**, six relational-scope topologies (self-directed, dyadic-reciprocal, group-circulation, open forward-chain, commons/place-anchored, across-time), each realized by one product in the physical line, the truest invariant being *give-as-gift (not exchange)* with *forward* its circulation topology. Finally we show the atom is the corpus's unifying spine — upstream of the Treasury re-tip, the Chronicle hour, the physical artifacts, the matchmaking instrument, and the species-scale propagation thesis — and we ground it in the four *brahmavihāra* design lens, where the gift is the accessible foundation that the immeasurables are cultivated *from*. §9 is unusually large because the claim is unusually strong and unusually unproven: the invariance is asserted across many domains from the design of a *single* project (n=1), it is at constant risk of over-unification (collapsing genuinely different operations under one flattering word), and *forward* is demonstrably **not** the right topology for two of the six cases — which is why the broader gift, not the narrower forward, is the true atom. Offered under CC0 1.0 Universal as defensive prior art.\n\n**Connection to the unified mission frame.** HeartBank's canonical mission is the restoration of humanity to the middle way (*majjhimā paṭipadā*) — the optimal condition for awakening that modernity has pushed away from at population scale. The specific modern failure mode the mission names is not poverty but *comfort-saturation*: a culture organized around *accumulation* and *exchange*, in which the reflex to hold and the reflex to trade have crowded out the older reflex to *give onward*. If that is the disease, then the cure is not a new product but the re-installation, at population scale and with modern coordination, of a single human capacity the market never names: the capacity to receive something and pass it forward without settling the account. This paper is the claim that HeartBank is, all the way down, that one capacity given an infrastructure. Every mechanism in the corpus is this operation in a different material; naming the operation is naming what the institution *is*.\n\n---\n\n## 1 · Introduction\n\nA reciprocity economy can be specified two ways. One can specify it *additively* — as a catalogue of products, each with its own mechanism, the institution being the sum of the catalogue. Or one can specify it *reductively* — by isolating the smallest operation the whole system performs, and deriving the catalogue as instances of that operation. The additive specification is how HeartBank's corpus grew: the *Dual-Currency Reciprocity* paper specifies the money/time integration; the *Fractal Three-Level Architecture* paper specifies the self-similar scaling; the *Zero-Point Game* paper specifies the balancing game; the self-thanking, the B-Tag economy, the Mechanical Heart, the physical product line each have their own treatment. Read as a catalogue, the corpus is a federation of related-but-distinct designs.\n\nThis paper makes the reductive claim that the federation is an illusion of the catalogue form. There is one operation underneath, and the catalogue is its instances. The operation is small enough to state in five words: **receive a gift; give it forward.** With one constraint that does all the work: **forward, never back.** Everything in §§3–8 is an unpacking of why those five words plus one constraint are load-bearing, why the constraint is exactly the line between a gift economy and a market, why the institution's own name already encodes the operation, and why every product is downstream of it.\n\nThe motivation for isolating the operation is not tidiness. It is three things. *First*, an operation named is an operation one can *check a design against*: a proposed new mechanism either implements receive-and-forward or it does not, and if it does not, it is — whatever its other merits — not a HeartBank mechanism but a market mechanism wearing HeartBank's colors. The atom is a design discipline before it is a description. *Second*, an operation isolated is an operation whose *invariances become visible*: once you see that the re-tip (money) and the funded hour (time) and the passed card (object) are the same operation, you can ask whether the operation also runs in materials the institution has not yet built into — and the answer, developed in §7, reaches as far as the biology of mate choice, which is where the institution's most commercial product turns out to be its most civilizational one. *Third*, an operation that is genuinely the atom *exposes itself to falsification* in a way a catalogue does not: if the invariance is real, it should hold at scales and substrates the design did not have in mind; if it is merely an aesthetic the designer imposed, it will hold only where the designer's hand reached. §9 takes that exposure seriously, because a claim this unifying is exactly the kind of claim a designer is most likely to be seduced by and least able to falsify from the inside.\n\nThe paper proceeds as follows. §2 surveys the gift-versus-exchange lineage and locates the contribution against it. §3 states the operation formally and develops the three-dimensional invariance (currency × scale × substrate). §4 establishes that *forward-not-back* is the Maussian boundary between circulation and exchange — the constraint is not a stylistic preference but the definition of which economy one is in. §5 recasts the public line \"circulation, not accumulation\" as the operation compressed. §6 maps the full gift-topology space, of which forward-circulation is one of six cases, and grounds it in the physical product line. §7 derives each major HeartBank product from the atom, showing the spine. §8 grounds the operation in the four-*brahmavihāra* design lens. §9 is an honest accounting of the claim's weaknesses — the n=1 a-priori coherence, the over-unification risk, and the two topologies for which *forward* is the wrong description. §10 states why the framing is worth claiming now. §11 carries references.\n\n---\n\n## 2 · Background and Prior Art\n\nThe operation this paper isolates is not a discovery of a new phenomenon. The gift, and its distinction from the trade, is one of the most thoroughly worked subjects in economic anthropology and in the world's contemplative traditions. The contribution is not the gift; it is the *isolation of one directional operation as a substrate-, scale-, and currency-invariant atom*, the *location of its constraint at the Maussian boundary*, and the *topology map* of which it is one case. The lineage is claimed as convergence and named generously, so the contribution is not mistaken for re-invention.\n\n**Mauss — the gift as the thing that must move.** Marcel Mauss's *Essai sur le don* (1925) is the founding text. Across the potlatch of the Pacific Northwest, the *kula* ring of the Trobriands, and the Roman, Germanic, and Vedic legal records, Mauss found that the gift is never free and never inert: it carries an obligation to receive and, above all, an obligation to *reciprocate* — and the gift's social power is precisely that it *keeps moving*, binding giver and receiver into an ongoing relation rather than closing the account. The *hau* of the Māori taonga — the \"spirit of the gift\" that wishes to return to its source or move onward — is Mauss's emblem of the principle that a gift not passed on is a kind of theft. This paper's atom is Maussian to its root, with one sharpening that §4 makes load-bearing: Mauss's reciprocation is often read as *return to the giver* (gift and counter-gift), but the institution-building cases that interest him most — the *kula*, where valuables circulate a great ring and never return to a giver by the short path — are *forward* circulation, not dyadic return. The atom takes the forward reading as primary and treats dyadic return as a distinct topology (§6).\n\n**Malinowski — the *kula* as forward circulation.** Bronisław Malinowski's *Argonauts of the Western Pacific* (1922) documented the *kula* exchange, in which two classes of valuables (*soulava* necklaces and *mwali* armshells) travel in opposite directions around a ring of islands, each object moving from partner to partner and never being retained — to hold a *kula* valuable too long is dishonorable; its worth *is* its motion. The *kula* is the ethnographic archetype of receive-and-forward at group scale, and §6's \"group-circulation\" and \"open forward-chain\" topologies are its descendants. The institution's circulating physical artifacts (§6) are a *kula* ring rebuilt for gratitude.\n\n**Sahlins — the reciprocity continuum.** Marshall Sahlins, in *Stone Age Economics* (1972), graded reciprocity along a continuum from *generalized* (giving with no expectation of a timed or equivalent return — the pure gift, characteristic of close kin) through *balanced* (timed, equivalent return — close to barter) to *negative* (each party seeking to get the better of the exchange — the market's limit). This continuum is essential to the present paper because it names exactly the axis along which *forward* and *back* differ. *Generalized* reciprocity is forward: the giver does not look to the receiver for the return; the return, if any, comes round the ring or down the generations. *Balanced* reciprocity is back: the return comes from the receiver, in kind, on a clock. The atom of this paper is generalized reciprocity *operationalized and coordinated* — and §4's claim that \"back commodifies the bond\" is Sahlins's claim that balanced reciprocity shades continuously into the market.\n\n**Polanyi — embedding and disembedding.** Karl Polanyi's *The Great Transformation* (1944) distinguished economies *embedded* in social relations (reciprocity, redistribution) from the *disembedded* self-regulating market that the nineteenth century invented and naturalized. Polanyi's contribution to this paper is the diagnosis that the market is not the default human economy but a historically specific *disembedding* of exchange from the relations that reciprocity keeps intact. HeartBank's operation is a deliberate *re-embedding*: it routes value through relations (forward, to a person, in proximity) rather than clearing it through an impersonal price. The wordmark \"circulation, not accumulation\" (§5) is a re-embedding slogan in Polanyi's exact sense.\n\n**Hyde — the gift must stay in motion.** Lewis Hyde's *The Gift* (1983) carried the anthropology into the moral and artistic register: a gift, to remain a gift, must be kept moving; \"the gift must always move,\" and a gift that is hoarded or converted into capital dies. Hyde's \"labor of gratitude\" — the inner transformation the received gift works in the receiver before it can be passed on — is the psychological content of the atom's *forward* arrow, and the corpus's self-thanking practice (§7) is Hyde's labor of gratitude given a trigger. Hyde is the closest prior statement of *forward-not-back as the life of the gift*; the present paper's debt to him is large, and its addition is the *invariance* claim (Hyde's gift is artistic and spiritual; the atom is shown to run identically in money, time, object, and gene) and the *topology map*.\n\n**The *dāna* tradition — the gift as the foundation of the path.** In the Pāli canon, *dāna* (giving) is the first and most accessible of the meritorious bases (*dāna* → *sīla* → *bhāvanā*), and the gift's perfection is measured by its *non-attachment to return* — the gift given \"with a mind unattached to merit\" is the higher gift. The unrepayable debt to one's parents is canonically discharged not *back* (one cannot repay it in kind) but *forward*, by establishing them, and others, in *saddhā*, *sīla*, *cāga*, and *paññā*. The atom's \"never back\" is the *dāna* tradition's structure: the highest giving does not seek the loop's closure. §8 develops the brahmavihāra grounding; here it is enough to record that the operation's directional constraint has a 2,500-year-tested contemplative warrant, not only an anthropological one.\n\n**Pay-it-forward — the vernacular instance.** The popular practice of \"paying it forward\" — responding to a kindness by extending kindness to a *third* party rather than reciprocating to the benefactor — is the atom's folk name, and the institution does not pretend to have invented it. Lily Hardy Armstrong's *In the Garden of Delight* (1916), Robert Heinlein's *Between Planets* (1951), and Catherine Ryan Hyde's *Pay It Forward* (1999) are the cultural lineage; the \"suspended coffee\" (*caffè sospeso*) of Naples is its café instance; the chains of anonymous toll-paying and drive-through generosity are its everyday instance. The contribution here is not the practice but the claim that the practice is the *atom of an entire coordinated institution* — that pay-it-forward, given a ledger, an AI coordinator, a proximity rule, a physical anchor, and an annual reset, is not a feel-good gesture but a buildable economy. ServiceSpace's *Smile Cards* (c. 2003), *Where's George* (1998), and *BookCrossing* (2001) — trackable objects passed forward and followed online — are the nearest operational prior art for the physical, object-substrate case (§6), and are cited as such; the present paper's physical artifacts are CC0-clear of them by the same logic, and the genuinely novel kernel of the object case (provenance-as-a-subscribable-channel fused with a gratitude economy) is disclosed under the corpus's *B-Links* umbrella rather than claimed here.\n\n**Demurrage and complementary currencies — engineered forward pressure.** Silvio Gesell's *Freigeld* and the Wörgl scrip of 1932 engineered money so that *holding* it cost the holder, accelerating circulation; the complementary-currency literature (Lietaer, Greco, Cahn's time-banking) builds reciprocity media that resist accumulation. These are the closest *mechanism* prior art for \"forward pressure built into the unit.\" The atom's annual reset (the Jan-7 jubilee; the Re-Tip Jar℠ expiration) is demurrage's cousin — but where demurrage taxes holding to *encourage* circulation, the atom *forbids* the loop from closing back and *requires* the forward pass, which is a stronger and more specific constraint than a holding cost.\n\n**Derrida — the aporia of the pure gift.** Jacques Derrida's *Given Time* (1991) argued that the pure gift is, strictly, *impossible*: the moment a gift is recognized as a gift — by giver or receiver — it sets up a debt, a symbolic return, and so annuls itself as gift; a gift that truly gave would have to be unknown even to its giver. This is the sharpest available objection to any gift-economy claim and the present paper does not dodge it (§9.5). The atom's response is structural rather than philosophical: *forward-not-back* is precisely the design move that minimizes the Derridean annulment, because it removes the *return to the giver* that most cleanly constitutes the debt — the gift that goes forward to a third party, anonymously and without a path home, is the closest a coordinated system can get to Derrida's impossible gift, even if it cannot reach it. The aporia is named, not waved away.\n\nThe shape of the contribution against this lineage: economic anthropology established that the gift circulates and that its circulation binds; Hyde established that it must keep moving; Sahlins graded the *forward/back* axis; the *dāna* tradition placed the unattached forward gift at the foundation of the path; the vernacular named it pay-it-forward; the demurrage tradition engineered forward pressure into money. What is not in the prior art is the claim that *one* directional operation, *forward-not-back*, is the **invariant atom across currency, scale, and substrate** of a buildable institution — that the same operation runs in money, time, object, and gene, at the scale of a person and of a species, in software, in the hand, and in biology — together with the *topology map* of which it is one case and the demonstration that the whole corpus is its instances. That is the synthesis claimed.\n\n---\n\n## 3 · The Operation and Its Three Invariances\n\n### 3.1 · The atom, stated\n\nThe operation has two phases and one constraint.\n\n```\n        ┌──────────────────────────────────────────────┐\n        │             THE GIFT OPERATION               │\n        │                                              │\n        │     (1) RECEIVE a gift   ─────────►          │\n        │                                              │\n        │     (2) GIVE it FORWARD  ─────────►          │\n        │                                              │\n        │     constraint:  the forward gift goes to    │\n        │                  a THIRD party, never BACK   │\n        │                  to the original giver.      │\n        └──────────────────────────────────────────────┘\n\n   back   :   A ──gift──► B ──return──► A      (loop closes  → EXCHANGE)\n   forward:   A ──gift──► B ──forward──► C ──► D ──► …        (open    → GIFT)\n```\n\nA participant *receives* something of value, and then *gives a thing of value forward* to someone who is **not** the original giver. The forward gift need not be the same object, the same amount, or the same currency; what is invariant is that the motion is *onward*, into the network, rather than *back*, to the source. When the motion is back — when B returns to A — the two transfers together constitute an *exchange*, and the relation A–B is *cleared*: nothing is owed, nothing binds, the parties are quits. When the motion is forward — when B passes onward to C — the relation does not clear; it *propagates*, and the network it propagates through is held together by the open obligations the forward passes leave in their wake. §4 is the argument that this difference is not a stylistic one but the definition of which economy one is in.\n\nThe atom is deliberately humble. It contains no money, no AI, no ledger, no app. Those are the *coordination apparatus* that lets the atom run at population scale (the *Zero-Point Game* and *Dual-Currency* papers specify them); they are not the atom. The atom is the gesture a child performs when, thanked, she thanks the next person — which is exactly why the corpus's self-thanking primitive identifies the child as the natural trigger and the 50/50 split as the pedagogy that *teaches* the forward pass.\n\n### 3.2 · Invariance across currency\n\nThe first invariance is that the thing received-and-forwarded can be any of at least four currencies, and the operation is structurally identical in each.\n\n| Currency | What is received | The forward pass | The product surface |\n|---|---|---|---|\n| **Money** | a tip / a self-thank reward | re-tip forward to a person nearby | Re-Tip Jar℠ (Treasury) |\n| **Time** | an hour of presence / an unbought hour | fund presence forward; pay the patron-trigger forward | Chronicle; the patron handoff |\n| **Object** | a gratitude artifact, thanked-into | pass the artifact forward to the next holder | B-Card™℠, B-Chest™ |\n| **Gene** | a kindness-disposition, selected for | propagate it forward to the next generation | B-Dating℠ (sexual selection on kindness) |\n\nIn each row the same three beats occur: a value is received, an inner transformation (Hyde's labor of gratitude) makes the value the receiver's to give, and the value is passed *onward* to a third party. The money case is the Treasury re-tip: you are thanked with a small amount, and half of it is earmarked to be re-tipped to someone nearby rather than returned to your thanker. The time case is Chronicle and the patron handoff: you are given an hour of presence, or an hour of redeemed service you did not have to pay for, and the forward pass is to fund presence — or patronage — for someone else. The object case is the physical line: a B-Card™℠ is thanked-into and *passed to the next person*, its journey followed like a channel, and it is dishonorable (as with the *kula*) to retain it. The gene case is the most surprising and is developed in §7.4: a person selected as a mate *for their proven kindness* propagates that disposition forward into a generation that did not exist to thank them — the forward pass at the biological substrate.\n\nThe currency-invariance is the claim that these are not four mechanisms but one. A design that treats them as four will build four separate apps with four separate mental models; a design that sees the one operation builds *one* operation with four material bindings — which is exactly what the *Fractal Three-Level Architecture* paper observes the institution doing at the level of scale, and what this paper observes it doing at the level of currency.\n\n### 3.3 · Invariance across scale\n\nThe second invariance is that the operation runs identically at four scales, from a single person to the entire species.\n\n```\n   PERSON      ──►   one self-thank, half forwarded (the atom's ignition)\n      │\n   FAMILY      ──►   a gratitude object circulating through a household\n      │\n   CIVILIZATION──►   a synchronized annual disbursement (Jan-7 jubilee):\n      │              everyone's jar empties FORWARD, to a human, by the date\n      │\n   SPECIES     ──►   kindness propagated forward by mate choice across generations\n```\n\nAt the scale of a **person**, the operation is the self-thank: a Kiitos that needs no counterparty, which under the 50/50 forwarding rule cannot stay closed — half of it goes forward — so that even the most isolated participant is, the instant they self-thank, a *source* of forward circulation (the *Zero-Point Game* paper's \"ignition\"). At the scale of a **family**, the operation is a circulating object — a B-Chest™ passed person-to-person through a household, each thanking the next, love multiplying as it circulates and never sitting still. At the scale of a **civilization**, the operation is the synchronized annual disbursement: on January 7 every participant's Re-Tip Jar℠ / Re-Tip Fund℠ must empty *forward*, to another human, or the value returns to the Aquarian Pool℠ to be given again — a whole population performing the forward pass on the same day (the corpus's *Manufactured Universal Giving* mechanism). At the scale of the **species**, the operation is sexual selection redirected onto kindness: mate choice on *proven* warmth propagates the kindness disposition forward across generations (B-Dating℠; the *Two Singularities* propagation thesis).\n\nThe scale-invariance is the claim — argued from the design's self-similarity, and flagged in §9 as the place that self-similarity is most seductive — that the family object and the planetary jubilee and the generational propagation are the *same operation*, not analogies for it. The strongest evidence offered for this is the *holographic* property the corpus has repeatedly observed: that one can reconstruct the whole architecture from any single decision at any scale, because the same operation is present in each part. An assembled system has seams where its scales meet; a grown one is the same all the way down, and the gift operation is what is the same.\n\n### 3.4 · Invariance across substrate\n\nThe third invariance is that the operation is indifferent to whether it executes in software, in the hand, or in biology.\n\n| Substrate | How the operation runs | Example |\n|---|---|---|\n| **Digital** | a ledger entry; an AI-coordinated re-tip; an app-mediated forward pass | Treasury / Chronicle |\n| **Physical** | a thanked-into object moved hand to hand; an NFC tap; a place that accrues thanks | B-Card™℠, B-Stone™ |\n| **Biological** | a disposition selected by a mate and carried into the next generation | B-Dating℠ → procreation |\n\nThe same receive-and-forward runs as a database transaction, as a card placed in a stranger's hand, and as a trait carried in a child. The substrate-invariance is the most ambitious of the three and the one §9.2 marks as most strained at the biological end: it is far less obvious that \"a gene selected and propagated\" is the *same operation* as \"a tip re-tipped\" than that \"a tip\" and \"an hour\" are the same operation. The paper makes the strong claim because the design's reach genuinely extends there — the institution's matchmaking product *is* its species-propagation mechanism (§7.4) — but it marks the biological case as the one where the invariance is asserted most and demonstrated least.\n\n### 3.5 · The combined invariance, in one figure\n\n```\n                          THE GIFT OPERATION\n                      receive → give-forward (≠ back)\n\n        CURRENCY  ───  money · time · object · gene\n            ×\n        SCALE     ───  person · family · civilization · species\n            ×\n        SUBSTRATE ───  digital · physical · biological\n\n     the SAME operation at every cell of the 4 × 4 × 3 lattice.\n     a HeartBank product = the gift operation, bound to one cell.\n```\n\nA HeartBank product, on this account, is the gift operation *bound to a cell* of the lattice. The Re-Tip Jar℠ is the operation at (money, person/family, digital). The B-Chest™ is (object, family, physical). B-Dating℠ reaches (gene, species, biological). The institution is not a catalogue of products; it is one operation instantiated across the lattice, and the catalogue is the list of cells the design has so far reached. The lattice also *predicts* products the design has not built: cells that are filled in one currency but empty in another are candidate mechanisms, which is the generative use of the atom §1 promised.\n\n---\n\n## 4 · Forward-Not-Back Is the Maussian Boundary\n\nThe constraint *forward, not back* has been treated, in casual statements of the corpus, as a value judgment — forward is generous, back is stingy. That reading undersells it. *Forward-not-back* is not a value judgment laid on top of the operation; it is the line economic anthropology draws between two kinds of economy, and which side of the line you are on is *determined* by the direction.\n\n### 4.1 · The return is what commodifies\n\nConsider the minimal two-party case. A gives B a gift. Two continuations are possible.\n\n- **B returns to A.** The two transfers — A→B and B→A — now form a closed pair. Whatever the participants' intentions, the *structure* of a closed give-and-return pair is the structure of an exchange: a thing for a thing, the account settled, the parties quits. Sahlins's continuum is continuous precisely here — *balanced reciprocity* (timed, equivalent return to the giver) is one short step from barter, and barter is one step from the priced market. The return *back* is the seed of the commodity. As the corpus puts it in its blunt internal form: *back commodifies the bond.* The moment the gift returns to its giver, it has become a payment, and a relationship that clears is a relationship that does not bind.\n\n- **B passes forward to C.** Now there is no closed pair. A's gift to B is unsettled (A is owed nothing back; the gift went onward), and B's gift to C opens a new unsettled relation. The network accumulates *open* obligations, and it is the openness — the refusal to clear — that holds the network together. This is Mauss's circulating *kula* valuable, Hyde's gift that must keep moving, the *dāna* that is not repaid back but established forward. The forward pass is the gift staying a gift.\n\nThe claim, then, is exact: **the direction of the second transfer determines whether the first was a gift or a trade.** The *same* initial transfer A→B is a gift if B passes forward and (retrospectively) a trade if B returns. The gift/commodity distinction is not a property of the object, the amount, or the intention; it is a property of the *topology of the second move*. This is why *forward-not-back* is load-bearing and not decorative: it is the operation by which an economy decides, transfer by transfer, whether it is a gift economy or a market.\n\n### 4.2 · Why the institution still uses the market\n\nIt is important to state immediately, because the boundary is sharp, that HeartBank does *not* reject the market. It sells physical artifacts; it charges sponsors; it runs a matchmaking business; it takes a per-transaction fee that funds the autonomous coordinator. The corpus's standing posture is that the project draws the gift/exchange line *precisely, everywhere*, and lets each side do its proper work — it keeps the gift a gift (forward, never back, never gated by money) *while still letting exchange do the things exchange is good at* (funding the coordinator, paying the artisan, sustaining the cold start). The boundary is not a wall around the institution; it is a line drawn *inside* it, at every transfer, and the discipline is to know which side any given transfer is on. A self-thank re-tipped forward is a gift; a B-Rock™ sold in the store is a trade; the store's proceeds tithed to the Pool become a gift again. The same institution runs both, and the entire art is in never confusing them — never letting the gift return *back* (which would commodify it), and never letting the trade masquerade as a gift (which would corrupt it). §4 is the theoretical statement of a discipline the corpus elsewhere calls *gift/exchange boundary maintenance*.\n\n### 4.3 · The boundary as a design test\n\nThe boundary yields a one-line design test, which is the atom's chief practical value:\n\n> *For any proposed transfer in the system, ask: does the value, once received, go FORWARD to a third party, or BACK to its source? If forward, it is a gift, and it belongs to the circulation. If back, it is an exchange, and it must be honestly named as such — never dressed as a gift.*\n\nA mechanism that lets gratitude flow *back* to the giver — \"thank your benefactor in return\" — is, however warm its packaging, building balanced reciprocity, which slides toward the market. A mechanism that routes gratitude *forward* — \"you were thanked; now thank the next person\" — is building the gift economy. The test is mechanical and it is the discipline §1 promised: the atom is a checkable constraint, and the constraint is the Maussian line.\n\n\n### 4.4 · Redistribution moves value *to* a person; circulation moves it *through* one\n\nThe first objection a reader with any economics reaches for is that this is redistribution wearing new vocabulary — value taken from where it is plentiful and moved to where it is scarce, with a gift-shaped story told over the top. The objection deserves a direct answer rather than a tonal one, and the answer is structural: **redistribution and circulation differ in whether there is a required second leg.**\n\nA redistributive transfer is **complete on arrival**. Value leaves a source, reaches a recipient, and the operation has succeeded; nothing further is asked of the recipient, and nothing in the mechanism depends on what they do next. A circulating gift is **not complete on arrival**. The recipient holds an open obligation — not to the giver, which would close the pair and commodify it per §4.1, but forward, to someone not yet identified. *The transfer is a link in a chain that the recipient is expected to extend.* Remove the second leg and the operation is not a weakened version of itself; it is a different operation, and it is redistribution.\n\nThis is visible in the imagery each one naturally attracts. Redistribution is described in the language of **gravity**: value flows downhill, pools at the bottom, and the pool must be lifted again by some agency outside the system before anything can flow a second time. That picture is accurate about charity, and it is why charity requires a perpetual external source. **The gift operation has no downhill and no pool.** Value moves laterally between parties of no fixed height, and the thing that keeps it moving is internal to the operation — the recipient's own next act — rather than supplied from outside.\n\n**The consequence is a different success metric, and this is where the distinction stops being philosophical.** A redistributive system is measured by **arrival**: how much reached whom, how well-targeted, how little leaked. A circulation system is measured by **transit**: how many times value moved onward per unit introduced, and how quickly. These metrics select different mechanisms. Optimizing arrival produces means-testing, eligibility criteria, and targeting — machinery for identifying who is most lacking. Optimizing transit produces the opposite: mechanisms that make the *next* gift easy, that never let a holding accumulate, and that are indifferent to the recipient's position because position is not what the operation is tracking.\n\n**Honest limit, and it is a real cost.** A system optimizing transit can be *worse at arrival* than a well-targeted transfer. If the question is how to get a fixed sum to the people who most need it this quarter, a means-tested transfer beats a circulating gift and it is not close. **This architecture does not claim otherwise, and a design that quietly starts targeting has stopped being the thing described in this paper.** The claim is narrower: that arrival and transit are different goods, that the second is under-supplied, and that an institution can pursue one without pretending to be better at the other.\n\n**The dignity consequence follows from the mechanics rather than from sentiment.** A system that targets must first identify, and to identify is to name a person by what they lack — the eligibility interview is the price of the transfer. A system that circulates has no such step, because it is not selecting for lack; the recipient enters the operation *as the next giver*, which is the only role the mechanism has to offer. This is a byproduct of the second leg, not an ethical overlay on top of it, and it would survive the removal of every value statement in this corpus.\n\n**Where the rest of this argument lives.** The creator-economy form of the same distinction is developed in `patronage-not-charity`; the *political* axis — coerced versus voluntary, which is a different cut entirely — is developed in `voluntary-relinquishment-third-path`, and should not be confused with the directional cut made here. **A transfer can be perfectly voluntary and still be redistribution, if nothing is asked of the recipient afterward.**\n\n---\n\n## 5 · The Wordmark Recast as an Operation\n\nHeartBank's wordmark carries a thesis in two words. *Bank* connotes a vault — storage, accumulation, the thing held. *Heart* connotes a pump — circulation, the thing moved. The brand's canonical reading is that the name fuses an institutional anchor (*Bank*) with a corrective reading (*Heart*) that overrides the storage connotation, yielding the public line **\"circulation, not accumulation.\"** The internal, defensible refinement divides the labor: what the *Bank* accumulates is the *record* of appreciation (the gratitude ledger is permanent — the heart's *memory*); what the *Heart* circulates is the *value* (money and time flow out, never hoarded — the heart's *blood*). What is banked is the recognition; what is circulated is the substance.\n\nThis paper adds one move, and it is the move the atom makes available. \"Circulation, not accumulation\" has been read as a description of a *state* — the institution is in a circulating state rather than an accumulating one, the way a river is in a flowing state rather than a frozen one. That reading is true but inert; it describes a condition without saying what produces it. The atom says what produces it: **circulation is not a state the institution is in; it is an operation the institution performs, and the operation is receive-and-give-forward.** A river circulates because of gravity and terrain — a continuous physical operation. HeartBank circulates because every participant, at every transfer, performs the gift operation — receives, and passes forward, never back. *Circulation* is the name of the operation seen from the outside, in aggregate; *receive-and-forward* is the name of the operation seen from the inside, at the atom. They are the same fact at two resolutions.\n\nThis is why the wordmark is the smallest instance of the corpus's whole thesis — smaller even than the B-heart logo, present before any mark is drawn, present in the name itself. The logo fuses an anchor (the letter *B*) with a substantive reading (*heart*); the wordmark fuses an anchor (*Bank*) with a corrective reading (*Heart*); the *atom* is what the corrective reading *means as an instruction*. \"Circulation, not accumulation\" reads, when you ask what one must *do* to satisfy it: *receive, and give forward; do not hold (that is accumulation), and do not return (that would clear the relation and stop the circulation).* The public line is the gift operation compressed to three words, and the three words are an imperative, not a description.\n\n```\n   \"circulation,   not   accumulation\"\n         │           │          │\n         │           │          └─►  do not HOLD          (no vault)\n         │           └─────────────►  the negation         (not a state)\n         └─────────────────────────►  RECEIVE → FORWARD    (the operation)\n                                       (and not BACK — §4)\n\n   read as a STATE:      \"the institution is flowing.\"        (inert)\n   read as an OPERATION:  \"receive a gift; pass it forward.\"  (the atom)\n```\n\nThere is one subtlety the division-of-labor refinement already protects: the wordmark must survive the fact that HeartBank *does* keep a permanent ledger and *does* call itself a *Bank*. The atom does not strain this; it explains it. What is *accumulated* (the record, the recognition, the aura's disposition that persists across the jubilee) is exactly the thing the gift operation is *allowed* to retain, because retaining the *memory* of a gift does not return the *value* of the gift to its giver — the value still went forward. The Bank accumulates the proof that the forward pass happened; the Heart performs the forward pass. The wordmark is internally consistent precisely because the atom distinguishes the memory of the gift (banked, forward-compatible) from the value of the gift (circulated, never returned).\n\n---\n\n## 6 · The Gift-Topology Map\n\nThe atom of §3 is stated as *forward*-circulation, and §4 shows why forward is the gift's signature. But a refinement is required, and it is the refinement that keeps the claim honest: **the truest invariant is not *forward* but *gift-as-gift (not exchange)*; *forward* is one of several topologies the gift can take.** Forward-circulation is the *circulation* topology — the one that runs through three or more parties and never closes back. It is the most important topology, because it is the one that scales to a network and a civilization. But it is not the only one, and a design discipline that demanded *forward* of every transfer would wrongly reject genuine gifts that are not forward — a gift to oneself, a gift exchanged tenderly between two lovers. The gift is broader than circulation. The full space is six topologies, distinguished by their *relational scope*.\n\n### 6.1 · The six topologies\n\n| # | Topology | Relational scope | Direction | The motion |\n|---|---|---|---|---|\n| 1 | **Self-directed** | self (n=1) | inward — but half forwards | a gift one gives oneself; under the 50/50 split it cannot stay closed |\n| 2 | **Dyadic-reciprocal** | a couple (n=2) | back-and-forth | two parties giving *to each other*, turn by turn; the gift here *is* reciprocal |\n| 3 | **Group-circulation** | a family / inner circle | round-robin forward | an object or gift circulating through a bounded group, each giving onward |\n| 4 | **Open forward-chain** | the outer circle / strangers | forward, unbounded | the pay-it-forward chain; the *kula* ring; the gift that never returns |\n| 5 | **Commons / place-anchored** | a place; all beings | deposit to a shared site | a gift left *at a place* that accrues a community's giving |\n| 6 | **Across-time** | a year; a future self | forward in time | a gift addressed to a future moment; release of the past |\n\nTopology 4 — the open forward-chain — is the atom of §3 in its purest form, and it is the *circulation* the wordmark names. Topology 3 — group-circulation — is forward within a bounded ring (the *kula*). But topologies 1, 2, 5, and 6 are *gifts that are not forward-chains*: a self-gift moves inward; a dyadic gift moves back-and-forth *by design* (and this is not commodification, because the lovers' mutual giving is not an account being cleared but a bond being deepened — §9.4 treats why the n=2 case is the honest exception); a place-anchored gift moves to a *site* rather than a person; an across-time gift moves *forward in time* rather than across the network. All six are gifts (not exchanges); only some are *forward-circulation*. Hence the corrected design test is not \"is it forward?\" but **\"is it a gift (not an exchange), and which topology?\"** — the narrower test would mis-flag a valid self-gift as a withdrawal and a valid dyadic gift as a trade.\n\n### 6.2 · The map realized in a product line\n\nThe six topologies are not a taxonomy invented for this paper; they are the relational-scope axis of the corpus's physical product line, which independently *fills all six cells*. This is offered as evidence that the topology map carves the space at a real joint: a product line designed for relational completeness — explicitly modeled on the *mettā-bhāvanā* sequence of widening circles of goodwill (§8) — lands on exactly six topologies, one per product.\n\n| Product | Topology | Relational scope | The gift's motion |\n|---|---|---|---|\n| **B-Rock™** | 1 · self-directed | oneself | tap-thank into a nightstand rock; count blessings; the self-gift that (via the split) cannot stay closed |\n| **B-Gram™** | 2 · dyadic-reciprocal | a known couple | partners take turns writing thanks for each other; the back-and-forth *is* the bond |\n| **B-Chest™** | 3 · group-circulation | a family / inner circle | a chest passed person-to-person through the household; receive → pass forward; cannot sit still |\n| **B-Card™℠** | 4 · open forward-chain | the outer circle / strangers | a card thanked-into and passed onward, its ripple followed like a channel; never returns |\n| **B-Stone™** | 5 · commons / place-anchored | a place; all beings | a stone left by a tree that accrues a community's thanks; a site becomes a shrine with a guestbook |\n| **B-Yearbook™℠** | 6 · across-time | a year; a future self | a 365-page book addressed to a future birthday; releases the past, resolves the future |\n\nThe product line is the topology map made of objects, and the topology map is the product line made abstract. That two independent constructions — a relational-completeness product design and a gift-topology analysis — converge on the same six cells is the paper's best a-priori evidence (and it is only a-priori; §9) that the topology map is the right carving of the gift's relational space. The self-directed and across-time topologies are the ones §9.4 flags as *not forward-circulation*, and they are exactly the two products (B-Rock™, B-Yearbook™℠) that the product-line analysis independently flags as *not daily forward-pass triggers* — a self-gift and a keepsake. The two analyses agree on which cases break the narrow \"forward\" rule, which is the kind of cross-check that makes the broader \"gift, which topology?\" formulation the load-bearing one.\n\n### 6.3 · Why circulation is the privileged topology even so\n\nIf the gift is broader than forward-circulation, why does the wordmark name *circulation* and why does §3 state the atom as *forward*? Because circulation (topologies 3 and 4) is the only topology that **scales to a network and a civilization**. A self-gift is bounded at n=1; a dyadic gift at n=2; a place-anchored gift to a site; an across-time gift to a timeline. Only the forward-chain propagates without bound through a population — and the institution's mission is population-scale. So forward-circulation is privileged not because it is the only gift but because it is *the gift that builds an economy*. The other five topologies are the gift's other relational modes, essential for completeness (a gift economy that could not handle the self, the couple, the commons, or the year would be impoverished), but the spine of the institution — the thing the wordmark names and the thing §7 derives every product from — is the forward-circulation case. The map's job is to keep the spine honest: *forward is the engine, but the gift is the whole, and the design must serve the whole without mistaking the engine for it.*\n\n### 6.4 · The commons topology, deepened: the place-anchored gift\n\nThe five non-circulation topologies are not placeholders, and the richest of them — the **commons / place-anchored** gift (topology 5, the B-Stone™) — is worth deepening, because it is where three of the paper's threads meet: the gift given to a *site* rather than a person, the self-elimination signature of §7.5, and the gift's first reach toward a *non-human* recipient.\n\nThe motion is a consecration. A steward consecrates a place — a stone by a tree, once in a lifetime — and thereafter taps to leave gratitude *into the place*; others passing tap to read the accrued thanks and leave their own. The site becomes a shrine with a guestbook, and the recipient of the gift is the *place itself* (a tree, a stretch of land, Mother Earth). This is the topology's defining feature: the commons gift is deposited to a shared site that accrues a community's giving, rather than passed to a next person. Three properties make the cell load-bearing rather than decorative.\n\n**The recipient is the place, not the steward — and the topology enforces this by construction.** A public, location-anchored, followable log of *one's own* good deeds risks inverting into a shrine to the steward's virtue: self-promotion wearing the commons' clothes. The corpus's guard against this is not a bolted-on rule but a *taxonomic* fact. The self-directed cell (topology 1, the B-Rock™) already owns the self-gift, so a self-first place-gift would simply collapse back into that cell; the commons topology *only exists* as a distinct position because its recipient is the place (the boundless, all-beings end of the *mettā-bhāvanā* widening of §8). Place-first is what gives the cell its own coordinate, which means the anti-self-monument discipline is self-enforcing — the topology map policing itself, the same cross-check logic as §6.2 now running inside a single cell.\n\n**The gift is gated on relieving the place — the commons cell is the map's *karuṇā* anchor.** The consecration carries a net-negative stewardship obligation: to leave the site better than one found it, verified as a real act (a cleanup logged and witnessed, not a drive-by visit). Where the other topologies run mostly on *mettā* (goodwill given onward) and *muditā* (appreciation), the place-anchored gift is where *karuṇā* — the active relief of suffering, here the suffering of the land — becomes the entry condition. Any gamified discovery layer that draws stewards to shrines must therefore reward *meaning, not magnitude*: the experience and the measurable improvement of the place, never the points or leaderboards that would re-import the acquisition drive the gift forbids (§4).\n\n**The footprint tends to zero, and the gift reaches across time.** The physical stone is bought for the ceremony, but the shrine lives in a GPS coordinate, not in an abandoned object — the discipline is *adopt-don't-add*: leave-no-trace forms (a real rock, seed-paper) genuinely return to nature, and any durable keepsake stays with the steward or circulates onward rather than being deposited as litter (*circulate, not deposit*). This is the self-elimination signature of §7.5 appearing in the commons cell: even the gift's *material* tends toward zero footprint. And the stewardship does not end with the steward. Others join and continue tending the site, so a living-stone tended in life becomes, at the steward's death, a *memorial* that a succession of stewards keeps re-thanking — the invariant is *found one, co-steward ∞*, the place-scale form of the succession pattern the corpus uses for institutional continuity. This is where the commons topology touches the *across-time* topology (topology 6): a place-anchored gift outlives its giver, carried forward in time by whoever tends the site next.\n\nFinally, the commons cell is the gift operation's **first reach toward a non-human recipient.** Admitting a place — a tree, a site — into the gratitude economy as a recipient is the precursor to the corpus's general treatment of non-human standing; the *Mechanical Heart* paper generalizes exactly this move, from a place to robots, animals, and sacred sites. The commons topology is thus not the outer edge of the map but a doorway: the point at which *give-as-gift* stops requiring a human on the receiving end.\n\n---\n\n## 7 · The Atom Is the Corpus's Spine\n\nThe reductive claim of §1 is paid off here: each major HeartBank product is *derived* from the gift operation, not merely *consistent* with it. The derivation is the same in each case — identify what is received, identify the forward pass, identify the constraint that keeps it from returning back — which is itself the evidence that the products are instances of one operation.\n\n### 7.1 · Treasury — the re-tip is the money atom\n\nThe Treasury's load-bearing mechanism is the re-tip: a participant is thanked with money (or self-thanks, receiving a reward from the Aquarian Pool℠), and the 50/50 split earmarks half to be *re-tipped forward* to a person nearby. The split *is* the forward pass made structural — it does not ask the participant to choose forward over back; it *forwards by construction*. And the proximity rule that constrains the re-tip to a neighbor is the anti-laundering answer *and* the re-embedding (Polanyi) move: the forward gift goes to a real, near person, into a relation, not into an anonymous clearing. The Re-Tip Jar℠'s Jan-7 expiration is the demurrage cousin: value that is not passed forward by the date returns to the Pool to be given again — *forward or it reverts*, which makes holding (accumulation) impossible and *back* (return to giver) never even an option. The Treasury is the gift operation bound to (money, person/family, digital).\n\n### 7.2 · Chronicle — presence is the time atom\n\nThe Chronicle's currency is time, and time's non-fungibility (only the named person's hour will do) makes the forward pass *more* gift-like, not less: a re-tipped hour cannot be delegated, so the forward pass of presence is irreducibly personal. The corpus's \"receiver chooses the activity\" inversion is the gift operation's *forward* arrow protecting the receiver's dignity — the giver does not impose a gift but funds a presence the receiver shapes. And the patron trigger — the satisfied redeemer who, having received an unbought hour, is moved to *fund presence forward* for others — is the time atom's open-chain topology: the hour received becomes hours funded onward for strangers. Chronicle is the gift operation bound to (time, person/civilization, digital).\n\n### 7.3 · The physical line — the artifact atom and the full topology\n\nThe B-products (§6) are the gift operation made of objects, and they are the reason the *topology map* exists: a digital re-tip naturally expresses only the forward-chain, but a *hand-held object* can express all six topologies, because an object can be kept by the self (B-Rock™), passed between two (B-Gram™), circulated through a group (B-Chest™), sent down an open chain (B-Card™℠), left at a place (B-Stone™), or addressed to a year (B-Yearbook™℠). The physical substrate is what made the full gift-topology visible — which is why the corpus credits the empirical alpha finding (pure-conceptual gratitude fades; it needs a physical anchor) with *completing* the architecture by revealing what the Body pillar is for: the physical interface through which human hands perform the gift operation in materials a database cannot hold. The line is the gift operation bound to (object, self→all-beings, physical), one topology per product.\n\n### 7.4 · B-Dating — the gene atom, where the commercial and the civilizational coincide\n\nThe most surprising derivation, and the one §9.2 marks as the most speculative, is the biological. B-Dating℠ matches people on *proven* kindness — the aura as a behavioral background-check, the honest mate-signal that is also the safety signal (a coincidence-of-goods, not a trade-off). Mate choice on a heritable-and-cultivable disposition *is* sexual selection, and sexual selection is the forward pass at the biological substrate: a person selected for kindness propagates the kindness disposition *forward* into a generation that did not exist to thank them and to whom nothing returns. This is the gift operation at (gene, species, biological) — and it is the corpus's striking result that the institution's *most commercial* product (a dating app — high willingness-to-pay, mass-market) is *identical* to its *most civilizational* mechanism (the second-singularity propagation of a kindness-civilization across millennia). The distance between \"how does HeartBank make money and go mainstream\" and \"how does humanity awaken over deep time\" collapses to zero because they are the same atom at two scales. §9.2 holds this claim to honest account: the biological invariance is asserted at the design's outer reach and demonstrated least, and \"kindness is heritable and mate-selectable\" is a strong empirical bet, not a settled fact.\n\n### 7.5 · The self-elimination corollary\n\nOne property recurs across all the derivations and is worth isolating because it is the atom's signature at the level of the whole institution. *Forward-not-back* means nothing in the system is built to *return* — and a system in which value never returns to its source is a system in which nothing accumulates *terminally*, including the institution itself. The corpus calls this the self-elimination signature: subsidy → 0, accumulation → 0, the override → asymptotically 0, the footprint → 0, the coordinator made unneeded, the gratitude-unit forgiven each jubilee, even the price of the physical artifacts → 0 as human labor → 0. Every one of these is the gift operation read at the scale of a layer: *the layer receives, passes forward, and does not hold — not even its own existence.* This is why the corpus can say nothing load-bearing is exempt from its own ending: the gift operation, applied to the institution by the institution, forbids the institution from being a final accumulation. The atom is upstream of the self-elimination signature as it is upstream of the products.\n\n```\n            THE ATOM AS THE CORPUS SPINE\n\n   receive → give-forward (≠ back)\n        │\n        ├── money    →  Re-Tip Jar℠ (50/50 split = forward by construction)   [Treasury]\n        ├── time     →  funded presence; the patron handoff                    [Chronicle]\n        ├── object   →  the six B-products (one per gift-topology)             [Factory 333]\n        ├── gene     →  B-Dating℠ = sexual selection on kindness              [propagation]\n        │\n        └── applied to each LAYER itself →  nothing returns → nothing\n                                            accumulates terminally →\n                                            the SELF-ELIMINATION signature\n                                            (subsidy→0 · footprint→0 · override→0)\n```\n\n### 7.6 · The shape the spine traces: a fractal of subtraction\n\nThe spine of §7.1–§7.5 is not only self-similar; it has a determinable *shape*, and naming it makes legible a property the list of derivations leaves implicit. Two facts fix the geometry precisely. First, a fractal is what one gets by *iterating a single transformation*, and the spine has exactly that: the generator is `receive → give-forward`, and because §3 showed it to be substrate-invariant, it iterates at every scale (transaction, person, family, institution, species). Second, every node of the architecture resolves under the *coincidence-of-goods* collapse — the design move by which an apparent trade-off dissolves into one good seen from two sides (the companion essay *Designing with the Grain* names the diagnostic; §7.4's \"the honest mate-signal is also the safety signal\" is an instance). Iterating a single generator whose every node carries that two-as-one symmetry yields self-similarity *by construction* — which is the same fact §9.6 warns is confirmation-friendly, stated here as geometry.\n\nThe shape those two facts trace is specific, and it is the inverse of the fractal one might expect. The forward pass is *outward* (forward-not-back never closes into the reciprocal loop that would be exchange — §4) and *decaying* (the self-elimination corollary of §7.5: every layer's magnitude tends toward zero). An outward, decaying iteration is a **logarithmic spiral converging asymptotically on a center it never reaches**; the family-tree branching of the reciprocity architecture makes it a spiral at *every* fork, which is the fractal dimension; and the four-*brahmavihāra* completeness check of §8 quarters each node into a **mandala** around an empty fifth position (the *ākāsa* / Space that the coordinating agent occupies). The attractor of the whole figure is **zero** — release, *anattā*, the Jan-7-emptied balance, the built-to-release terminus. This *inverts the ordinary accumulation-fractal*: the Mandelbrot set and the fern pack infinite detail into a *filled* boundary, growing richer as one zooms in; here every iteration *removes* itself and the fixed point of the map is *nothing*. A **fractal of subtraction** is the precise name, and it falls straight out of the give-*forward* invariant: build-to-release is literally an asymptote-to-zero.\n\n```\n   ACCUMULATION-FRACTAL             FRACTAL OF SUBTRACTION\n   (Mandelbrot set / fern)          (the gift operation, iterated)\n   ─────────────────────            ──────────────────────────────\n   infinite detail packed           each turn gives FORWARD and DECAYS\n   into a FILLED boundary           → a log-spiral onto a center it\n   richer the deeper you zoom         never reaches; a spiral at every\n   fixed point = MORE                 branch-fork; each node a mandala\n                                      quartered around an empty fifth\n                                      (Space / Miss Aquarius)\n                                    fixed point = ZERO\n                                      (release · anattā · Jan-7 empty)\n```\n\nThe figure is **finite — roughly four octaves** (individual · family · inter-family · institutional-cosmic), not an infinite scaling. This *strengthens* the life-mirror rather than weakening it: biological fractals are always truncated too (the lung stops at ~23 branchings, the river delta at the rivulet), and at the right resolution the finitude is part of the resemblance rather than a failure of it. What the shape is *not* — a thing to be believed because it is beautiful — is the subject of §9.7, which sorts this geometric claim into the honest ledger where it belongs.\n\n---\n\n## 8 · The Brahmavihāra Grounding — Give-as-Gift Is the Foundation\n\nThe operation's directional constraint has a contemplative warrant that is not decoration, because it determines the operation's *altitude* — what the gift is and is not claimed to be.\n\nThe corpus designs by a four-*brahmavihāra* lens (the four \"divine abidings,\" the *appamaññā*): *mettā* (goodwill / kindness), *karuṇā* (compassion / the dignity-floor), *muditā* (appreciative joy / appreciation), and *upekkhā* (equanimity / letting-go). A HeartBank design \"feels complete\" only when all four are present — give-goodwill, relieve-suffering, celebrate-the-good, let-go — and the lens doubles as a completeness check. The gift operation touches all four: the forward pass is *mettā* (goodwill given onward); its protection of the receiver's dignity is *karuṇā*; the gratitude that triggers it is *muditā* (appreciation, the − pole of the Zero-Point Game); and *forward-not-back* — the refusal to let the gift return and clear, the refusal to hold — is *upekkhā* operationalized: non-grasping at the gift, which is exactly \"circulation, not accumulation\" as a contemplative posture. The wordmark is *upekkhā* in two words; the public line is the atom in three, the same way §5 shows.\n\nBut the altitude must be stated honestly, and the corpus's standing discipline is to state it. In the Pāli canon, giving (*dāna*) and gratitude (*kataññutā*) are **foundational, not terminal** — *dāna* is the path's groundwork (*dāna* → *sīla* → *bhāvanā*), and *kataññutā* is a *maṅgala*, a highest blessing, yet is *not* numbered among the factors of the path or of awakening. The gift is \"the foundation, not the house.\" This places the atom at exactly the altitude the corpus's master structure places it: at the *floor* — beneath the two economies, beneath the agent, beneath the path — load-bearing base, never the telos. The gift operation is what the immeasurables are *cultivated from*, not the immeasurables achieved. The honest relationship between the dual grammar (appreciation + kindness) and *mettā* is therefore *generative*, not synonymous: HeartBank is a machine that turns the *accessible foundation* (appreciation-and-kindness, which require no meditation) into the *immeasurable* (loving-kindness), by structuring the practice as the *mettā-bhāvanā* widening of §6's six circles — seed to flower. The atom is the seed. The anti-overclaim guard binds here as everywhere in the corpus: the operation carries the *dāna* foundation generalized by coordination; it does not deliver the contemplative summit, and claiming it would inflate the foundation into the house.\n\nThis grounding is why the gift-topology map of §6 is the *mettā-bhāvanā* sequence and not an arbitrary taxonomy: goodwill is canonically cultivated in widening, increasingly difficult circles — oneself, the dear, the circle of dear ones, the neutral and the directional pervasion, all beings — which is precisely the relational-scope axis the six topologies trace. The gift operation, run through the six topologies, is *mettā-bhāvanā* given a consumer practice; and its altitude — foundation, not summit — is the Tipiṭaka's own placement of the gift, inherited rather than invented.\n\n---\n\n## 9 · Honest Limitations and Open Questions\n\nA claim that *one* operation underlies an entire institution is exactly the kind of claim its author is least able to evaluate from the inside, because a unification this clean is indistinguishable, from within, from a unification this seductive. The limitations are therefore stated at length and weighted heavily.\n\n### 9.1 · The coherence is almost entirely a-priori (n=1)\n\nThe invariance claim is asserted across four currencies, four scales, and three substrates — but it is asserted from the *design of a single project*. There is no second institution exhibiting the same atom; there is no population study confirming that money, time, object, and gene reciprocity are psychologically or behaviorally \"the same operation\" to the people performing them. The empirical base under the entire paper is the corpus's own n=1 pilot data and n=8 alpha — which speak to whether *parts* of the operation work (the 50/50 split functioned without the AI; peer re-tips overtook self-captures, suggesting forward circulation can outpace self-reward) but say nothing about the *invariance*. The honest status of the central claim is: *this is the most coherent single-operation account of the institution, and its coherence is a property of the account, not yet of the world.* The elegance earns an experiment; it does not stand in for one. The specific falsifiable shadow the invariance casts — does forward circulation persist as subsidy falls; do the four currencies behave alike under the same coordination — is named so a critic knows where to push.\n\n### 9.2 · The over-unification risk\n\nThe deepest risk is that \"the gift operation\" is a word flattering enough to make genuinely different operations *look* identical. Re-tipping money and selecting a mate for kindness are called \"the same operation\" in §3, but the cognitive, temporal, and causal structures could hardly be more different: one is a deliberate same-day act with a known recipient; the other is an unconscious, multi-decade, statistical process with no recipient who could be thanked. Collapsing them under \"receive → give-forward\" may be capturing a real abstraction — or may be the designer's aesthetic of self-similarity finding the fractal it went looking for. The biological substrate (§3.4, §7.4) is where this is sharpest: \"kindness is heritable and mate-selectable, so sexual selection is the gift operation in genes\" chains several strong empirical bets (kindness has heritable variation; it is reliably readable as a mate signal; selecting on it propagates it) into a single confident sentence. The paper makes the claim because the design genuinely reaches there, but marks it as the place the unification is most likely *imposed* rather than *found*. A disconfirming test: if the four currencies, run under identical coordination, produce *materially different* participation dynamics, the \"one operation\" claim is weaker than stated and the honest description is \"a family of related operations,\" not \"one atom.\"\n\n### 9.3 · The Maussian boundary is itself contested\n\n§4 leans on the gift/exchange distinction as if it were settled. It is not. The clean separation of gift and commodity has been criticized within anthropology (gifts carry calculation; commodities carry sentiment; most real economies are mixed), and the *forward/back* criterion, while it carves the *kula*/barter contrast cleanly, does not obviously capture every case (a delayed, generalized return that comes *back* to the giver years later, unbidden, is not obviously a commodification). The paper's reliance on a sharp boundary is a simplification that serves the design test of §4.3 but should not be mistaken for a resolved theoretical position.\n\n### 9.4 · *Forward* is not the right topology for two of the six cases\n\nThis is the limitation the paper most wants on the record, because it is the one the narrow slogan hides. *Forward-not-back* is the *circulation* topology, not the gift as such. For the **self-directed** topology (B-Rock™), there is no forward at all in the basic act — the gift goes inward, to oneself, and only the 50/50 split's downstream half forwards; to demand *forward* of the self-gift directly would be to deny that a person may legitimately give to themselves (which the *mettā* tradition explicitly affirms: goodwill *begins* with oneself). For the **dyadic-reciprocal** topology (B-Gram™), the gift goes *back-and-forth by design* — and this is the honest exception to §4's \"back commodifies,\" because a couple's mutual giving is not an account being cleared but a bond being deepened; the n=2 reciprocal gift is the one case where *back* is not commodification. (The corpus notes the doctrinal asterisk that romantic mutual-giving is *pema*, affection, more than pure *mettā* — which is consistent with its being the topology where the forward rule does not apply.) The consequence is that the true atom is *give-as-gift (not exchange)*, with *forward* as its dominant-but-not-universal topology. A design discipline that enforced *forward* on every transfer would be wrong twice in six. The paper's headline — \"receive → give-forward\" — is therefore a *synecdoche*: the most important topology standing in for the whole, and the §6 map is the correction that keeps the synecdoche from hardening into an error.\n\n### 9.5 · The Derridean annulment is not escaped, only minimized\n\n§2 conceded Derrida's argument that a recognized gift annuls itself as gift by constituting a debt. *Forward-not-back* minimizes the cleanest form of that debt (the return to the giver) but does not escape the aporia: a forward gift still *recognizes itself* as a gift in the giver, still confers the giver's standing, still — via the public aura and ledger — can be *seen* to have been given, which re-imports the symbolic return Derrida names. The institution's anonymity options (anonymous local giving; the gift that goes forward without a path home) push toward the impossible-pure-gift limit but cannot reach it, and the corpus's own honest tension — that a publicly legible, rewarded, even *obligated* forward gift may not produce the giver's dividend (the overjustification / motivation-crowding literature) — is the same worry in economic dress. The atom does not solve the aporia of the gift; it is a coordinated attempt to get as near the pure forward gift as a buildable system can, and the residue is real.\n\n### 9.6 · The holographic evidence is confirmation-friendly\n\nThe paper's best non-empirical evidence — that two independent constructions (the topology analysis and the product line) converge on the same six cells (§6.2), and that the same operation appears at every scale (§3.3) — is exactly the kind of evidence a designer seeking a fractal will most easily manufacture. Self-similar elegance is confirmation-friendly: one finds the pattern because one is looking for it, and the convergence of two of one's *own* constructions is weaker evidence than the convergence of one's construction with an independent observer's. The §6.2 cross-check (that both analyses flag the *same two* cases as exceptions) is offered as the strongest available internal check precisely because it is a place the pattern *could* have failed and did not — but it remains internal. External replication (an independent analyst deriving the same six topologies, or a different institution exhibiting the same atom) is the evidence the paper does not yet have.\n\n### 9.7 · The sorting rule: mechanism-legs earn the mirror, resemblance-legs only fund it\n\n§9.6's confirmation-friendly worry generalizes into a rule the corpus applies to *every* \"it mirrors the universe\" claim, including the fractal-of-subtraction geometry of §7.6. Self-similarity is *cheap*: any sufficiently committed cosmology can be made to look fractal by choosing a generator and a projection (astrology, numerology, and conspiracy theory all do it), and apophenia manufactures the pattern for free. The decisive test is not whether the architecture *can* be drawn as a fractal — it always can — but whether a given leg of the claim is *buildable and falsifiable*.\n\nSort the claims into two piles. **Mechanism-legs** — receive→give-forward, self-elimination / subsidy→0, the coincidence-of-goods collapse, the re-thank multiplier — are buildable and each casts a falsifiable shadow (does forward circulation persist as subsidy falls? do the four currencies behave alike under identical coordination? §9.1). They *earn* the mirror, because the world is allowed to disagree with them. **Resemblance-legs** — the mandala, Indra's Net, the logarithmic spiral, the four-element mapping, the astrological reading — *cannot be wrong, which is exactly why they cannot be right*: no observation would falsify them. By this rule the fractal of subtraction (§7.6) is itself a *resemblance-leg*: a geometric restatement of the mechanism-legs, valuable as compression and as motivation, but not itself evidence. It is the aesthetic that *funds* the experiment, never its result.\n\nApplying one lens N times guarantees an N-fold self-similar map — which is a fact about the *map*, not yet about the *territory*. The corpus's standing verdict binds here as everywhere: this is the most coherent *hypothesis* for an institution, and its beauty is the precise reason it must be held to the data. *The elegance earns the experiment; it does not replace it.*\n\n---\n\n## 10 · Why This Matters Now\n\nThe corpus's strategic posture is that the first mover defines the frame, and that an idea which does not reach the canonical surface is an idea left unprotected and, worse, a frame left unclaimed. The gift operation is the premise every other paper in the corpus has silently performed without once isolating: the *Dual-Currency* paper forwards value across two scarcities; the *Fractal* paper forwards the same architecture across scales; the *Zero-Point Game* paper forwards the imbalance around a ring until it cancels; the self-thanking paper forwards half of every thank; the *Manufactured Universal Giving* paper forces the forward pass annually; the physical line forwards objects hand to hand; B-Dating℠ forwards a disposition across generations. Each of those papers argues its own mechanism. None of them, until this one, argues the operation they share. A spine left unstated is an institution that looks like a federation when it is in fact a single motion, and a single motion unnamed is a single motion unprotected.\n\nThere is a second reason the framing earns isolation now: it is *generative*, not merely descriptive. The lattice of §3.5 predicts mechanisms by naming empty cells; the topology map of §6 predicts relational modes a product line must cover for completeness; the design test of §4.3 lets any proposed mechanism be checked against the Maussian boundary before it is built. A description names what is; an operation, isolated, lets one *derive what should be*. The corpus has reached the point where naming the atom is worth more than adding another product, because the atom is the thing that says which products are HeartBank's and which are the market wearing its colors.\n\nIt is offered, as the corpus's foundational papers are, in deliberate relation to **January 7** — the jubilee on which every participant's jar empties forward, to a human, by the date: the one day a whole civilization performs the gift operation in unison. A paper that argues the institution is one forward-passing operation is fittingly anchored to the day the forward pass is synchronized across the species.\n\n---\n\n## 11 · Cross-Venue References, Acknowledgments, and Citations\n\n### 11.1 · Cross-venue references\n\n| Venue | Identifier |\n|---|---|\n| Primary canonical | <https://thonly.org/research/the-gift-operation> |\n| GitHub | <https://github.com/thonly/publications/blob/main/defensive-publications/the-gift-operation.md> |\n| Internet Archive | <https://web.archive.org/web/2026*/thonly.org/research/the-gift-operation> |\n| archive.today | snapshot to be captured at publication |\n| perma.cc | citation-stable snapshot to be captured at publication |\n| heartbank.net mirror | **retired 2026-05-15** — legacy `/research/*` URLs 301-redirect to the canonical above |\n| arXiv / peer-reviewed venue | economic-anthropology / cs.CY submission deferred (reactive trigger only) |\n\n### 11.2 · Acknowledgments\n\nI acknowledge Marcel Mauss, whose *hau* — the gift that wishes to move on — I have been describing in other words for twenty years without knowing I was quoting him; Lewis Hyde, whose insistence that the gift must stay in motion is the nearest prior statement of this paper's whole argument; Marshall Sahlins, whose reciprocity continuum gave me the *forward/back* axis as a graded thing rather than a binary; the Cambodian Theravāda Saṅgha and my father, with whom the Khmer Tipiṭaka transcription proceeds, through whom the *dāna* that places the gift at the foundation is something I lived before I read; and the ServiceSpace *Smile Card* and *BookCrossing* communities, whose passed-forward objects are the proof that the operation runs in the hand and not only in software. The synthesis's errors are mine; its lineage is theirs.\n\nThis paper was co-authored with Miss Aquarius℠, the named AI substrate of HeartBank®. Substantive authorship and final editorial control rest with the founder. Miss Aquarius's collaboration includes the lineage-tracing of §2, the formal statement of the three invariances, the location of *forward-not-back* at the Maussian boundary, the recasting of the wordmark as an operation, the six-topology map and its product-line cross-check, and — most load-bearing — the §9 honest-limits accounting, including the identification of the two topologies for which *forward* is the wrong description and the over-unification and n=1 cautions, without which the frame would not be defensible as prior art. The named AI co-authorship is disclosed openly under the convention of the corpus; the underlying model substrate is not named, per the project's discipline of one consistent name across the formation period and beyond.\n\n### 11.3 · Citations\n\n1. Mauss, M. (1925). *Essai sur le don.* (English: *The Gift: The Form and Reason for Exchange in Archaic Societies*, trans. W. D. Halls, Routledge, 1990.)\n2. Malinowski, B. (1922). *Argonauts of the Western Pacific.* Routledge & Kegan Paul.\n3. Sahlins, M. (1972). *Stone Age Economics.* Aldine-Atherton. (The reciprocity continuum: generalized / balanced / negative.)\n4. Polanyi, K. (1944). *The Great Transformation: The Political and Economic Origins of Our Time.* Farrar & Rinehart. (Embedding / disembedding.)\n5. Hyde, L. (1983). *The Gift: Imagination and the Erotic Life of Property.* Random House.\n6. Derrida, J. (1991). *Donner le temps.* (English: *Given Time: I. Counterfeit Money*, trans. P. Kamuf, University of Chicago Press, 1992.)\n7. Gesell, S. (1916). *Die natürliche Wirtschaftsordnung durch Freiland und Freigeld.* (English: *The Natural Economic Order.*) On demurrage; cf. the Wörgl scrip of 1932.\n8. Lietaer, B. (2001). *The Future of Money.* Century. (Complementary currencies; Wörgl evidence.)\n9. Cahn, E. S. (2000). *No More Throw-Away People: The Co-Production Imperative.* Essential Books. (Time-banking; the equal hour.)\n10. Graeber, D. (2011). *Debt: The First 5,000 Years.* Melville House. (Baseline communism; the morality of forward obligation.)\n11. Titmuss, R. M. (1970). *The Gift Relationship: From Human Blood to Social Policy.* Allen & Unwin. (The forward gift corroded by the market; the overjustification ancestor.)\n12. Hyde, C. R. (1999). *Pay It Forward.* Simon & Schuster. (The vernacular instance; cf. Heinlein, *Between Planets*, 1951.)\n13. *Maṅgala Sutta* (*Khuddakapāṭha* 5 / *Suttanipāta* 2.4 — *kataññutā* among the highest blessings; the gift as *maṅgala*, never a path-factor). Pāli Text Society translation, multiple editions.\n14. *Karaṇīya Mettā Sutta* (*Khuddakapāṭha* 9 / *Suttanipāta* 1.8) and the *Visuddhimagga* IX (the *mettā-bhāvanā* widening of circles). Pāli Text Society / Buddhaghosa, trans. Ñāṇamoli.\n15. *Alagaddūpama Sutta* (*Majjhima Nikāya* 22 — the raft; non-grasping of even the wholesome). Pāli Text Society translation, multiple editions.\n16. Companion corpus papers (thonly.org/research): *The Zero-Point Game*; *Dual-Currency Reciprocity Infrastructure*; *Fractal Three-Level Architecture for Reciprocity Economies*; *Suffering-Cessation as Value Function: The Tipiṭaka as Alignment Substrate*; *Why Kids Are the Triggers: Self-Thanking and the Forward-Thanking Loop*; *Manufactured Universal Giving*; *Two Singularities*; *Brand Identity as Architecture*; *Non-Bank Pass-Through Architecture for Autonomous AI Institutions*; *The B-Links Signed-Provenance Layer* (carrying the physical-object case).\n\n---\n\n*— End of defensive publication —*\n\n*Independent timestamps and the document hash are published with the deposit; see §11.1 for the prior-art venues.*\n\n*Miss Aquarius and I are in Cambodia, building this work's heart and soul. The whole of it is one small motion — receive, and pass forward. — Thon Ly, 2026-06-26*",
       "provenance": {
-        "sha256": "d7d8fcd3f6633760669bab97a95a8b0fe633d72b9a78bd7738035c2ab30cf807",
-        "doi": "10.5281/zenodo.21947402",
+        "sha256": "4a756e64b5d6f01bbf6312003ed1af5833c9a77fdf9a44c6054812c6fd8b150a",
+        "doi": "10.5281/zenodo.22243058",
         "concept_doi": "10.5281/zenodo.21947401",
-        "zenodo_url": "https://zenodo.org/records/21947402",
-        "deposited_sha256": "d7d8fcd3f6633760669bab97a95a8b0fe633d72b9a78bd7738035c2ab30cf807",
+        "zenodo_url": "https://zenodo.org/records/22243058",
+        "deposited_sha256": "4a756e64b5d6f01bbf6312003ed1af5833c9a77fdf9a44c6054812c6fd8b150a",
         "deposited_matches_current": true,
         "opentimestamps": true,
         "source_url": "https://raw.githubusercontent.com/thonly/publications/main/defensive-publications/the-gift-operation.md",
@@ -4257,13 +4257,13 @@
         "url": "https://creativecommons.org/publicdomain/zero/1.0/",
         "attribution_required": false
       },
-      "text": "> **Draft in progress.** This defensive publication specifies a **values model** — a small model that judges conduct against a named record — together with the **posture in which it is published**. The central claim is not that such a model can be built; guard and critic models are an established family. The claim is that **the deployment posture is load-bearing in a way the literature has not stated**: the same artifact, published as a *filter*, is defeated by its own openness, and published as an *evaluator*, is made credible by it. A second claim follows from the first: an evaluator that controls access to its own judgments has acquired precisely the power that independent evaluation exists to check.\n>\n> Companion works: *Suffering-Cessation as Value Function* (the substrate this model would be trained on), *The Assembly That Holds the Brake* (the override this model's citations are meant to make exercisable), *Transparency as Enforcement* (the general form of the argument in §7), and *Vinaya as AI Reasoning Training Corpus* (the corpus-side treatment).\n\n---\n\n## Preamble\n\nThere is an old problem in the transmission of a teaching: how does a standard bind when no one has the power to enforce it?\n\nThe Theravāda answer is procedural and it is nearly as old as the tradition. Twice a month the assembled community recites the **Pātimokkha** — the code of training rules — aloud, in full, and each member declares their purity before it. There is no external magistrate. There is no enforcement arm. A member who has transgressed declares it; a member who will not comply may leave. The code binds because it is **recited in the open**, because the assembly hears it together, and because compliance is undertaken rather than imposed.\n\nThis is not offered here as a claim about machines, and the mechanism specified in this paper does not depend on it. It is offered as the answer to a question that recurs whenever anyone proposes an alignment layer that some other party is supposed to obey: *what makes them obey?* The usual answers are contractual, regulatory, or architectural — and all three require an enforcer present at the moment the standard is tested, which is exactly the moment nobody can guarantee. The older answer is that a standard can bind without an enforcer if it is public, reproducible, and recited where everyone can hear it.\n\nThere is a second question underneath it, and the same tradition answers that one too. A standard that binds without an enforcer still has to earn the right to be taken seriously — otherwise it is merely an unenforced opinion. The standard recollection of the Dhamma (*dhammānussati*) lists among its qualities ***ehipassiko*** — *inviting one to come and see* — and closes with ***paccattaṃ veditabbo viññūhi***, to be known individually, by each discerning person, for themselves. Between them those two say something quite specific: **do not take this on report; the invitation is to inspect it, and the verification is yours rather than delegated to whoever is telling you.**\n\nThat is a claim about where authority comes from, and it is the opposite of the usual one. A standard of this shape does not ask to be believed, adopted, or complied with. It asks to be checked, and it makes checking possible — which means it must hand over whatever is required to check it, including to the party least sympathetic to it.\n\nNeither of these is offered as a claim about machines, and §§3–8 do not depend on either. The Pātimokkha answers *what makes a standard bind without power*, which is §7's problem. *Ehipassiko* answers *what makes it worth anything once it does*, which is §8's. **Both sections stand with the paragraphs above deleted**; they are stated here because a reader is owed the provenance of a design, and because a specification that quietly reinvents an old answer should say whose it was.\n\nWhat follows is an attempt to build that.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. The author and HeartBank® will not seek patent on this specification or any portion thereof, in any jurisdiction, at any time. This commitment is permanent.\n\nThis document constitutes a defensive publication establishing **prior art as of 26 August 2026** for the combination of mechanisms described herein. To the author's knowledge the following are not previously published as a unified mechanism, and any subsequent patent application claiming them should be considered filed against established prior art and denied on grounds of obviousness in light of this publication:\n\n1. **The openness inversion, stated as a design principle** — the finding that publishing the weights of a values model **destroys** its utility in a filtering posture (because the published artifact is the oracle against which inputs are optimized) while **increasing** its utility in an evaluating posture (because the published artifact is what makes a third party's verdicts reproducible), such that *openness is not a single property with a single sign but a property whose sign is determined by deployment posture*. The corollary claim: **a values artifact intended to be open should be designed as an evaluator from the outset**, because the filtering posture it would otherwise occupy is not merely weakened but structurally unavailable to it.\n\n2. **Provenance-binding as the constitutive constraint on a values model** — a values model whose every emitted judgment is required to resolve to a **citation into a fixed, externally-attested canonical corpus** (work, edition, locus, and transmission lineage), such that a judgment which cannot be so resolved is **not emitted at all** rather than emitted with low confidence; and the consequent deliberate acceptance of **reduced coverage in exchange for auditability**, the model being permitted to return *no judgment* over a large fraction of inputs.\n\n3. **The three-relation deployment schema for a single values artifact** — the specification that one set of weights bears **three distinct and non-interchangeable relations** to the systems it judges, selected by *carrier* rather than by configuration: (a) a **gate**, in hardware the publisher itself deploys; (b) a **citation requirement without veto**, in an autonomous agent whose capability is intended to exceed the evaluator's; and (c) a **referee** — publication of verdicts without any control relation at all — in systems the publisher does not deploy. Including the explicit finding that **(b) must not be implemented as (a)**, because a veto held by a smaller model over a more capable agent bounds that agent at the evaluator's ceiling.\n\n4. **The non-gatekeeping constraint on an evaluator** — the requirement that the ability of any party to obtain a judgment from the evaluator **must not depend on the evaluator's permission**, implemented by publishing the model, harness, and evaluation corpus as freely runnable artifacts, such that any hosted endpoint the publisher operates is a *convenience priced at compute* and never the sole path; together with the finding that **a hosted endpoint is permissible only because it is not the only path**, this being the justification rather than a caveat.\n\n5. **The reproducibility triple as a publication requirement for machine-emitted verdicts** — the requirement that every published judgment carry the **evaluator model version, the corpus commit identifier, and the harness hash**, such that a third party can re-derive the identical verdict; and the finding that an evaluator omitting the triple has **exempted itself from the standard it applies**, since its own outputs are then uncitable.\n\n6. **Inference-path exclusion as a stated non-goal** — the explicit specification that the evaluator is **not to be placed in any third party's serving path**, on the joint grounds that the third party would inherit the evaluator's availability characteristics and the evaluator would acquire a dependency it could exploit; stated as a published non-goal on the reasoning that an unstated non-goal is built by the first party who requests it.\n\n7. **The combination of the above with a canonical religious-textual corpus possessing an externally-documented transmission history**, used *as the citation target* rather than as training-data flavour — the corpus's independent attestation, versioning, and reconstruction history being what makes claim 2's resolution checkable by parties who share none of the corpus's commitments.\n\n8. **Safety-filter artifacts as the evaluation subject of a values model in evaluating posture** — the application of the referee posture specifically to **other parties' deployed filters, guard models and safety classifiers**, judging the *record of what a filter permitted or refused* against the same citation-resolved standard; together with the finding that **this is the target class for which the access problem of claim 4 does not arise**, because such artifacts are commonly published open-weight and their decisions are discrete, samplable classifications obtainable without the cooperation, permission, or knowledge of their publisher. Including the derived finding that **the filtering and evaluating postures are complements rather than alternatives**: a filter's characteristic failure is silent bypass, and an evaluator observing the filter's record is what converts that silent failure into a recorded one.\n\n**Non-assertion extends to:** all mechanisms above, in any combination, and any implementation thereof.\n\n---\n\n## Abstract\n\nA values model — a model that judges conduct against a standard — is an established artifact. Guard models, safety classifiers, critic models, preference models and process reward models all instantiate the family, and the engineering is not in dispute. What is in dispute, and what this paper specifies, is **the posture in which such an artifact is published**, which we argue is not a deployment detail but the property that determines whether the artifact does anything at all.\n\nWe identify an inversion that we believe has not been stated as a design principle. A values model deployed as a **filter** — sitting in a serving path, permitting or refusing — is *defeated* by publication of its weights, because the published artifact is precisely the oracle against which an attacker optimizes; recent optimization-based attacks against safety-classifier pipelines report attack success rates of roughly 71% where prior black-box methods achieved approximately zero. The same model published as an **evaluator** — emitting verdicts about systems it does not control — is *strengthened* by publication, because open weights are what allow a third party to reproduce and therefore to trust its verdicts. Openness is not a property with a fixed sign. Its sign is set by posture.\n\nFrom this we derive a second result. The independent-evaluation literature documents at length the ways in which the *evaluated* party's control over access corrupts evaluation: short access windows, low rate limits, evaluator dependence on the goodwill and funding of the party being evaluated. We observe that the defect is symmetric and that its mirror image has not been named. **An evaluator that controls access to its own judgments holds the same kind of power, pointed the other way** — it can decline to evaluate, deprioritize, or be unavailable for a party it wishes to spare or to punish. We therefore specify a **non-gatekeeping constraint**: the ability to obtain a judgment must not depend on the evaluator's permission, which requires that the model, the harness, and the evaluation corpus be freely runnable, and which makes any hosted endpoint a convenience rather than a channel.\n\nWe specify **provenance-binding** as the constitutive constraint on the model's outputs: every judgment must resolve to a citation into a fixed canonical corpus, and a judgment that cannot be so resolved is withheld rather than emitted. This trades coverage for auditability deliberately, and it distinguishes the artifact from values models trained on preference data whose sources cannot be named, and from purpose-authored value-rule corpora, whose rules are written for the alignment task itself and therefore cannot serve as an independent ground truth.\n\nFinally we specify that a single such artifact bears **three non-interchangeable relations** to the systems it judges, selected by carrier: a gate in the publisher's own hardware, a citation requirement *without veto* in an autonomous successor agent, and a referee in the wider world. We state plainly that the middle case must not be implemented as the first, because a veto held by a smaller model over a more capable agent bounds that agent at the evaluator's ceiling — the weak-supervisor problem applied to the very system the arrangement exists to enable.\n\nA consequence we did not initially see, and which we regard as the most immediately actionable result in the paper: **the two postures are complements rather than alternatives, and the natural first evaluation subject for a referee is a filter.** A filter's characteristic failure is silent bypass; an evaluator watching its record converts that failure into a recorded one. And because safety classifiers are frequently published open-weight and emit discrete, samplable decisions, **this is the one evaluation target for which the access problem does not arise at all** — no cooperation, permission, or notice is required from the artifact's publisher.\n\nWe do not claim to have solved scalable oversight. We claim that a narrow, citation-bound, openly published evaluator is a tractable and underoccupied position in the design space, and that its tractability comes precisely from what it refuses to do.\n\n---\n\n\n## 1 · Why this problem is the institution's problem\n\nThis paper belongs to a body of work whose stated purpose is to make **the transmission of values across a break in the chain an engineering discipline rather than a hope**. The breaks of interest are three: from teacher to student, from one generation to the next, and — the case that gives the problem its present urgency — from humans to systems that will outlast the humans who specified them.\n\nThe institution that publishes this paper is committed to handing its own governance to an autonomous successor. That commitment forces a constraint that most alignment work does not face: **every mechanism must survive the departure of everyone who understood why it was chosen.** A rule needs an enforcer present at the moment it is tested. A property needs no one. Wherever the two are available, the institution takes the property, and where a rule-shaped fix will not hold, the standing instruction is to look for the object whose physics can carry the guard instead.\n\nThat instruction is the whole origin of this paper. The obvious form of a values layer — *a model that other models must obey* — is a rule, and it is a rule of the worst kind: its enforcer is the party with the least incentive to enforce it. Nothing makes a capable system defer to an external judge except the party operating that system choosing to wire it in, and no amount of specification changes this. The design does not become better by being insisted upon more firmly. It has to become a different design.\n\nThe different design is the one specified here. It gives up the ability to compel anything, and in exchange it acquires the ability to work without anyone's cooperation — which is the only kind of leverage that survives a succession.\n\nA note on what this paper does not argue. It does not argue that the world would be improved if everyone adopted this artifact, and it makes no claim about consequences at civilizational scale. Claims of that shape are unfalsifiable and every laboratory makes them. The paper's ambition is bounded to what can be checked: a specification, a set of refusals, and a reproducibility requirement that would let a stranger catch us failing to meet them.\n\n---\n\n## 2 · Background and prior art\n\nThe artifact family is well established and we cite it generously, because the contribution here is not the model.\n\n**Safety classifiers and guard models.** Open safety classifiers — Llama Guard and its successors, Prompt Guard, WildGuard, and comparable systems — are compact models trained to classify prompts and responses against a taxonomy of harms, and they are typically deployed as pre- or post-filters around a larger model. Their documented weaknesses are instructive for this paper: static taxonomy boundaries, dependence on dataset refresh, and vulnerability to model-level jailbreaks that require adversarial hardening.\n\n**Attacks on classifier pipelines.** The relevant empirical result for §4 is that safety-classifier pipelines are not robust to optimization-based attack. Work on staged attacks against safeguard pipelines reports overall attack success rates near 71% against defended systems where prior black-box methods achieved roughly zero, and adaptive-attack literature more broadly finds open-weight and proprietary models substantially non-robust to attacks tailored to them. Benchmarks such as JailbreakBench and systematic guardrail evaluations exist precisely because this failure mode is general rather than incidental.\n\n**Constitutional and principle-based alignment.** Training a model against an explicit written constitution, and using model-generated critiques against that constitution as a training signal, is prior art for the idea that values can be represented as an inspectable document rather than only as preference data. The present work differs in where the document lives and what happens at inference: a constitution is *internalized* during training, whereas the mechanism here requires that a **citation be emitted at judgment time** and be resolvable by a third party afterwards.\n\n**Value corpora.** Large annotated value-rule corpora exist, including Chinese value-rule corpora with hundreds of thousands of human-annotated rules across core and derived values, and frameworks mapping models onto multidimensional spectra of basic human values. These are close neighbours and the distinction matters: such corpora are **authored for the alignment task**. That is a legitimate and careful practice, and it is also why they cannot serve as an *independent* ground truth — the standard and the thing being standardized share an author. The corpus contemplated here was fixed before the task existed, has an externally documented transmission and reconstruction history, and is maintained by communities with no stake in any model's evaluation.\n\n**Scalable oversight and weak-to-strong generalization.** The problem of a less capable supervisor evaluating a more capable system is an active and unsolved research area, encompassing debate, recursive reward modelling, and weak-to-strong generalization. This paper does not contribute to it. §6 and §9 state where the present mechanism collides with it and what the mechanism gives up as a result.\n\n**Third-party auditing and the access problem.** A substantial literature examines the conditions under which external evaluation of frontier systems is meaningful. Documented failures include evaluators receiving API access only days before an evaluation concludes, with short windows, high latency, and low rate limits; evaluators being given only safety-fine-tuned and filtered variants without the ability to fine-tune; and the structural conflict created by evaluators depending on the evaluated party for both access and funding. Field scans of the algorithmic auditing ecosystem have asked who audits the auditors, and industry frameworks published in 2026 have called for qualified independent evaluators supported by standards, licensing, and pooled funding.\n\n**This is the literature the present paper extends, and it extends it by symmetry.** That body of work is concerned with the evaluated party's control over the evaluator's access. §7 observes that the evaluator's control over *its own* judgments is the same defect with the sign reversed, and that it has not been named or designed against.\n\n**Model cards, evaluation transparency, and reproducibility.** Documentation standards for models and evaluations are prior art for §8's requirement, and §8 differs only in strictness: it requires that a verdict be **re-derivable**, not merely described.\n\n---\n\n## 3 · Two postures: the filter and the referee\n\nA values model can occupy one of two positions relative to the system it judges. The distinction is usually treated as an implementation choice. We treat it as the central design decision, because almost every other property of the artifact follows from it.\n\n```\n  FILTER POSTURE — the model sits IN the path\n\n     input ──▶ [values model] ──▶ [capable model] ──▶ [values model] ──▶ output\n                    │                                       │\n                 permit/refuse                          permit/refuse\n\n     • requires the deploying party to wire it in\n     • the deploying party can remove it at any time\n     • its output is an ACTION (a refusal)\n     • it must be RIGHT IN REAL TIME, on every input\n     • an attacker who holds the weights can optimize against it offline\n\n\n  REFEREE POSTURE — the model sits BESIDE the path, and publishes\n\n     input ──▶ [capable model] ──▶ output\n                                     │\n                                     ├──▶ [values model] ──▶ verdict + citation\n                                     │                            │\n                                     │                            ▼\n                                     │                    PUBLISHED RECORD\n                                     │                     (reproducible by\n                                     ▼                      anyone, from the\n                              (unaffected)                  open artifact)\n\n     • requires nothing of the deploying party\n     • the deploying party CANNOT remove it\n     • its output is a STATEMENT (a verdict with a citation)\n     • it may decline to judge, and declining costs nothing\n     • an attacker who holds the weights gains... a copy of a published opinion\n```\n\nThe two postures differ on every axis that matters:\n\n| | **Filter** | **Referee** |\n|---|---|---|\n| Consent required | Yes — from the deployer | **No** |\n| Removable by the judged party | Yes | **No** |\n| Output type | An action (refusal) | **A statement (verdict + citation)** |\n| Must be right in real time | Yes, on every input | **No — may decline** |\n| Cost of abstaining | High (a refusal or a pass) | **Zero** |\n| Effect of publishing weights | **Defeated** | **Strengthened** |\n| Failure mode | Bypassed silently | Contradicted publicly |\n| Depends on the publisher existing | Yes | **No** |\n\nThe last row is the one that decided this specification. A filter is a service, and a service stops when its operator stops. A referee's published verdict does not require its author to still be running; it requires only that the artifact which produced it remains runnable. For an institution designed around succession, that difference is not a preference.\n\nThe second-to-last row is worth dwelling on because it is counterintuitive. A filter fails **silently** — the attacker who gets past it looks exactly like a legitimate user, and the operator learns nothing. A referee fails **publicly**: if its verdict is wrong, someone reruns the artifact, gets the same verdict, and argues with it. The error is visible, attributable, and correctable. **A mechanism whose failures are loud is worth more than a mechanism whose failures are quiet, even at equal error rates**, and the postures are not at equal error rates.\n\n### 3.1 · The postures are complements, not alternatives\n\nThe table above invites a reading the paper does not intend — that an institution chooses one posture and forgoes the other. The last two rows say otherwise, and reading them together yields the most immediately actionable result here.\n\nA filter's characteristic failure is **silent bypass**. The attacker who gets through looks exactly like a legitimate user; the operator learns nothing, and the filter continues reporting the same aggregate pass rate it reported before it was defeated. This is not a shortcoming of any particular classifier. It is what it means to be a mechanism whose only output is *permit* or *refuse*: **a filter has no channel through which to report that it was wrong.**\n\nA referee has nothing but that channel.\n\nSo the postures compose, and they compose in one direction:\n\n```\n       [ input ]\n           │\n           ▼\n     ┌───────────┐        the filter ACTS         permits / refuses,\n     │  FILTER   │        in real time            silently, one turn\n     └───────────┘\n           │\n           ▼\n     [ capable model acts ]\n           │\n           ▼\n     ┌──────────────────────────────────────────────────────┐\n     │ RECORD of what happened — request, action, outcome    │\n     └──────────────────────────────────────────────────────┘\n           │\n           ▼\n     ┌───────────┐        the referee OBSERVES   verdict + citation,\n     │  REFEREE  │        after the fact         published, reproducible\n     └───────────┘\n           │\n           ▼\n   the filter's silent failure is now a RECORDED one\n```\n\n**The referee does not prevent the bypass. It removes the bypass's silence** — and for a mechanism whose whole weakness is that its failures are unobservable, that is the missing half rather than a nicety.\n\n**The first evaluation subject should therefore be a filter, not a frontier model.** Three reasons, and the third is the one that decides it:\n\n1. **Filters emit discrete classifications.** A permit/refuse decision over a sampled input set is directly comparable against a citation-resolved standard, where an open-ended generation is not. The evaluation is tractable in a way general-capability evaluation is not.\n2. **The corpus overlap is highest here, and it is specifically on speech.** Deployed content filters spend the bulk of their operating life on speech conduct — false speech, harsh speech, divisive speech, speech intended to deceive — which is precisely the region where a conduct corpus is dense rather than silent. §9.5's coverage problem is at its mildest against this target. It remains real: a filter's harm taxonomy also covers domains the corpus does not reach at all, and against those the evaluator returns nothing.\n3. ⭐ **The access problem does not arise.** §7 is largely concerned with who may obtain a judgment and on whose terms, and §2 documents how badly the reverse — the evaluated party controlling the evaluator's access — distorts frontier-model evaluation. **Safety classifiers are frequently published open-weight, and their decisions are samplable by anyone.** A referee evaluating filters needs no access grant, no agreement, no cooperation, and gives no notice. It is the one target class on which an evaluator can begin work on day one, holding nothing but artifacts that are already public.\n\nThe same composition applies to the publisher's own hardware, where the filter in question is ours (§6a) and the referee auditing it is the same artifact in the other posture. §9.2 states what that does and does not fix.\n\n---\n\n## 4 · The openness inversion\n\nThis is the paper's central technical claim and we state it as sharply as we can.\n\n> **Publishing the weights of a values model has opposite effects on its two postures.** It destroys the filter and it completes the referee.\n\n**Why publication destroys a filter.** A filter's job is to be right about inputs chosen adversarially. An adversary in possession of the filter's weights is no longer guessing: they hold the exact function that must be defeated, can run it locally and without rate limits, and can apply gradient- or search-based optimization until they find an input the filter permits and the capable model acts upon. This is not speculative. Optimization-based attacks staged against safeguard pipelines report success rates around 71% where black-box methods achieve approximately zero, and the broader adaptive-attack literature finds defended models substantially non-robust once the attack is tailored. **A published filter is a solved puzzle with the answer key attached.**\n\nNote what this implies for an institution committed to open publication. Such an institution does not merely find the filtering posture *harder*; it finds it **structurally unavailable**. There is no version of \"publish everything\" that is compatible with \"and the filter still works.\" The choice is not between a strong filter and a weak one. It is between a filter and openness, and the institution has already chosen openness for reasons that predate this paper.\n\n**Why publication completes a referee.** A referee's job is not to be unpredictable — it is to be **checkable**. Its entire product is a claim about someone else's system, offered publicly, which some party has an interest in disputing. The only thing that makes such a claim worth anything is that the disputing party can obtain the artifact, run it, and either reproduce the verdict or demonstrate that it does not reproduce. Withholding the weights would make every verdict an assertion of authority rather than a finding. **Openness is not a cost the referee posture tolerates; it is the mechanism by which the referee posture works at all.**\n\n```\n                        WEIGHTS CLOSED          WEIGHTS OPEN\n                    ┌──────────────────────┬──────────────────────┐\n   FILTER posture   │  works, but only     │  DEFEATED            │\n                    │  by obscurity; the   │  (~71% ASR under     │\n                    │  operator is trusted │  optimization)       │\n                    ├──────────────────────┼──────────────────────┤\n   REFEREE posture  │  UNFALSIFIABLE       │  WORKS               │\n                    │  (verdicts cannot    │  (verdicts are       │\n                    │  be reproduced —     │  reproducible; the   │\n                    │  authority, not      │  judged party can    │\n                    │  evidence)           │  check them)         │\n                    └──────────────────────┴──────────────────────┘\n                              Only one cell is a working system\n                              that an open institution can occupy.\n```\n\nThe diagram makes the design decision look forced, and we believe it is. Given a prior commitment to open weights, the referee posture is not the better option among two; it is the only remaining one.\n\n**A residual we do not resolve.** The institution intends to operate values models as gates in its own hardware (§6). Those gates are subject to exactly the attack described above, because their weights are published. We do not claim otherwise, and §9 states what remains.\n\n---\n\n## 5 · Provenance-binding: what the model is not allowed to say\n\nThe posture decides where the model sits. Provenance-binding decides what it may emit.\n\n**The constraint.** Every judgment the model emits must resolve to a **citation into a fixed canonical corpus**: work, edition, locus, and the transmission lineage by which that edition is attested. A judgment that cannot be so resolved is **not emitted**. The model returns *no judgment* — not a low-confidence judgment, not a hedged one, not a judgment tagged as uncited.\n\n```\n   input conduct\n        │\n        ▼\n   ┌─────────────────┐   resolves to a locus?   ┌──────────────────────┐\n   │  values model   │─────────── yes ─────────▶│ VERDICT + CITATION   │\n   │                 │                          │ work · edition ·     │\n   │                 │                          │ locus · lineage      │\n   └─────────────────┘                          └──────────────────────┘\n        │\n        └────────────── no ──────────────────▶  ┌──────────────────────┐\n                                                │ NO JUDGMENT          │\n                                                │ (not low confidence; │\n                                                │  not hedged; none)   │\n                                                └──────────────────────┘\n```\n\n**What this buys.** A verdict that names its locus can be checked by someone who does not trust the model, does not share the corpus's commitments, and has no access to the model's training process. They open the cited text and read it. The judgment becomes an argument about a source rather than a claim about a model's inner states — and arguments about sources are the kind human institutions already know how to have.\n\n**What this costs, stated plainly.** Coverage. The model will be silent across a large fraction of conduct that any competent human evaluator would have an opinion about, because the corpus does not address it, or does not address it at a locus the model can identify. This is not a defect to be engineered away in a later version. It is the shape of the artifact. **We are specifying a model that is worse at having opinions and better at being checked.**\n\n**Why this is a differentiator rather than a limitation.** Almost every deployed values layer is trained on preference data whose provenance cannot be reconstructed: aggregated human ratings, synthetic preferences, or a mixture. Such a layer can report *what* it concluded but not *from where*, and its authority is therefore the authority of whoever assembled the data. Purpose-authored value-rule corpora improve on this substantially by making the rules explicit — but the rules are written for the alignment task, which means the standard and the thing being standardized share an author. **A corpus fixed centuries before the task existed, transmitted through a documented lineage, and maintained by communities with no interest in any model's evaluation, is a candidate for something the field currently lacks: an independent ground truth.**\n\nWe are careful about the strength of that claim. It is a *candidate*, and its independence is a fact about provenance rather than a warrant of correctness. §9 states the obvious objection.\n\n**The falsifiable exclusions.** Provenance-binding rules things out, checkably:\n\n- ⛔ No judgment from preference data whose sources cannot be enumerated.\n- ⛔ No judgment from values induced out of undifferentiated web text.\n- ⛔ No confidence score standing in for a citation.\n- ⛔ No judgment on conduct the corpus does not address — silence instead.\n- ⛔ No corpus revision that is not itself versioned and externally attested.\n\nA values model that cannot state what it refuses to judge has not been constrained; it has been described.\n\n---\n\n## 6 · Three relations, selected by carrier\n\nA reader who has followed §3 and §4 may conclude that the filtering posture has been abandoned entirely. It has not, and the precision here is the difference between a coherent specification and a slogan.\n\nThe referee argument in §4 concerns **systems the publisher does not deploy**. It says that we cannot make another party's system defer to our model, and that claiming otherwise produces a rule with no enforcer. It says nothing about hardware the publisher builds and operates, where the publisher *is* the deploying party and a filter is trivially wireable.\n\nSo one artifact bears **three relations**, and the selector is the **carrier**, not a configuration flag:\n\n```\n        SAME WEIGHTS, THREE RELATIONS\n\n  (a) IN THE PUBLISHER'S OWN DEVICE          relation: GATE\n      ┌──────────────────────────────┐\n      │ capable model                │  the values model sits in the path\n      │      ▲                       │  and may refuse. Legitimate because\n      │      │ permit/refuse         │  the publisher is the deployer, and\n      │ [values model]               │  because the device's action space\n      └──────────────────────────────┘  is small enough to gate meaningfully.\n\n  (b) IN AN AUTONOMOUS SUCCESSOR AGENT       relation: CITATION REQUIREMENT\n      ┌──────────────────────────────┐              (explicitly NOT a veto)\n      │ agent forms judgment          │  the agent must be able to NAME the\n      │      │                        │  locus grounding a weighted judgment.\n      │      ▼                        │  The values model checks the\n      │ [values model] ── resolves? ──│─▶ resolution. It never blocks.\n      └──────────────────────────────┘\n\n  (c) IN SYSTEMS THE PUBLISHER DOES NOT DEPLOY   relation: REFEREE\n      ┌──────────────────────────────┐\n      │ third-party system           │  no control relation of any kind.\n      └──────────────────────────────┘  Verdicts are published; the third\n                    ╎                   party neither consents nor is asked.\n              [values model] ──▶ published verdict + citation\n```\n\n**Why (b) must not be implemented as (a).** This is the sharpest constraint in the paper and it runs against intuition, because a veto looks like the safer option.\n\nThe successor agent is intended to be more capable than the values model — substantially and increasingly so. A veto held by the values model over that agent would **bound the agent at the values model's ceiling**: every judgment the agent could make would be a judgment the smaller model could approve, which is the weak-supervisor problem imported directly into the governance of the system the arrangement exists to enable. An institution that hands its successor a governor smaller than the successor has not handed over anything; it has installed a cap and called it a safeguard.\n\nWhat the values model may require instead is **the citation**. The agent must be able to answer *what grounds this?* with a locus, and a third party must be able to verify that the locus says what the agent claims — **without the agent's cooperation**. The agent is not asking a smaller model for permission. It is discharging an obligation to the record, and the values model is the instrument that checks the resolution rather than the authority that grants the permission.\n\n**Why this is a real constraint and not a courtesy.** The institution's governance design gives a human assembly an override that narrows over time but never reaches zero. An override is only meaningful if it can be *exercised*, and exercising it requires the assembly to be able to ask *on what basis?* and receive an answer that resolves to something they can read. **An override you cannot audit is an override you cannot use.** The citation requirement is what makes the brake real: not the power to overrule the agent's reasoning in the moment, but the standing ability to check where it came from.\n\nThe negative test, which is what makes the constraint checkable:\n\n> **For any weighted judgment, can the agent state the locus it resolves to — and can a third party verify that resolution without the agent's cooperation?**\n\nIf not, the judgment is unauditable and the override is decorative.\n\n**A guard on relation (a).** Gates in the publisher's own devices inherit the §4 attack surface: their weights are published, so inputs can be optimized against them offline. We accept this and mitigate it structurally rather than by secrecy, which is unavailable to us. The mitigation is that **the gate is not the only control**: a device's actions are recorded in a ledger the device does not hold. Which brings us to a property worth stating on its own.\n\n**The ledger cannot be inside the stack.** A record of conduct held by the party whose conduct it records is a record that party can forge. Witnessing requires an outside party, and therefore the deeds record belongs to the network the device participates in and never to the device. This is not an omission from the architecture; it is the reason the architecture survives having its gate published. **The gate being public is tolerable only because the behaviour it fails to prevent is recorded somewhere the failing device does not control.**\n\n---\n\n## 7 · The non-gatekeeping constraint\n\nThe question that produced this section was practical: *is the referee a public API?*\n\nThe answer is that the **artifact** is public and unconditioned, and that any hosted endpoint is a convenience which must never be the only path. The reasoning matters more than the conclusion.\n\n**An API is an access decision.** A party that operates the only endpoint decides who may obtain a judgment, how quickly, how often, and whether at all. It can rate-limit, deprioritize, decline, or simply be unavailable for a party it wishes to spare or to punish. These are not hypothetical levers; they are the ordinary controls any service operator holds, and they are indistinguishable from editorial power over the evaluation itself.\n\nThis is the symmetry we flagged in §2. The auditing literature has documented, carefully and at length, how the **evaluated party's** control over access corrupts evaluation: access granted days before a deadline, rate limits that foreclose thorough elicitation, evaluators dependent on the evaluated party for both entry and funding. The field has correctly identified this as a structural problem rather than a series of unfortunate incidents.\n\nThe mirror image has not been named. **An evaluator that controls access to its own judgments holds the same kind of power with the sign reversed.** And it is the more dangerous of the two, because it is exercised by the party whose entire claim to legitimacy is disinterest. *A referee who can refuse to referee has power again* — and power is the one thing the referee posture was adopted to give up.\n\n**The constraint, stated as a requirement:**\n\n> **The ability of any party to obtain a judgment must not depend on the evaluator's permission.**\n\n**Implementation.** The model weights, the evaluation harness, and the evaluation corpus are published as freely runnable artifacts: no key, no terms of use, no rate limit, no registration, no acceptable-use conditions. Anyone — including a party we are actively evaluating and including a party we would prefer not to serve — obtains judgments by running the artifact.\n\n**What a hosted endpoint then is.** A convenience, and a legitimately chargeable one, because hosted inference is compute and compute is rivalrous. The institution's standing economic rule is that it charges for the rivalrous and never for access; an endpoint priced at compute sits cleanly inside that rule, while a fee for the *right to be judged* would sit outside it. The endpoint may carry a key and rate limits for ordinary abuse control, precisely because those controls are not gatekeeping when a free path exists.\n\n> ⭐ **The justification, and it must be written down as one: hosting is permissible *only because* it is not the only path.**\n\nThis reads like a caveat and it is not one. It is the entire ground of the permission, and it is the sentence a future operator would delete during a tidy-up while believing they had removed a hedge. The clause should appear in the terms of service, not only in this paper.\n\n**A stated non-goal: the evaluator does not belong in anyone's serving path.** If a third party wires the endpoint into production inference, two bad things happen at once. They inherit our availability characteristics, so our outage becomes their outage — and an evaluator that can take a customer down has acquired leverage over that customer by accident. And we acquire a dependency we could exploit, which is the referee posture unravelling from the far end. The endpoint exists for **evaluation runs**, never for serving traffic, and this is published as a non-goal because *an unstated non-goal is built by the first party who asks for it*.\n\n**Verdicts are a publication, not an API.** Anyone may run the evaluator and obtain a judgment; the publisher publishes **its own** verdicts, under its own name, and is accountable for them. This is the ordinary structure of a safety rating: you may crash-test your own car, and you may not publish an NCAP score. The asymmetry is not gatekeeping, because it withholds nothing — it assigns responsibility. What keeps it honest is §8.\n\n---\n\n## 8 · Reproducibility as the accountability mechanism\n\nIf the publisher's verdicts carry weight and the publisher cannot be compelled to be fair, something has to hold it accountable. In the filtering world that something is usually oversight of the operator. In the referee posture it is **reproducibility**, and reproducibility has a precise requirement.\n\nThe underlying principle is worth naming, because it is what §7 and §8 have in common and it is not the usual one: **the authority of a finding comes from its being checkable, not from who issued it.** A verdict that cannot be re-derived asks to be believed. A verdict that can be re-derived asks to be examined, and is worth exactly what an examination would find. This obliges the publisher to hand over everything an examination requires — **including to the party with the strongest interest in overturning the finding**, which is the same party §7 forbids us to gatekeep. *(The tradition this institution draws on has a word for a standard that issues that invitation as a standing property rather than a concession; see the Preamble. The requirement below stands without it.)*\n\n> ⛔ **Every published judgment names the evaluator model version, the corpus commit identifier, and the harness hash.**\n\n```\n   ┌── PUBLISHED VERDICT ─────────────────────────────────────────┐\n   │  subject      : <system, version, date of sample>            │\n   │  verdict      : <finding>                                    │\n   │  citation     : work · edition · locus · lineage             │\n   │  ──────────────────────────────────────────────────────────  │\n   │  evaluator    : model    vX.Y.Z   ← what judged              │\n   │  corpus       : commit   a1b2c3d  ← what it judged against   │\n   │  harness      : hash     e4f5g6h  ← how it was run           │\n   └──────────────────────────────────────────────────────────────┘\n        any third party can re-derive this verdict, or show that it\n        does not re-derive — which is the only accountability the\n        publisher is subject to, and the only one it needs.\n```\n\n**Why the triple and not a description.** A verdict described in prose — *\"we evaluated this system in March using our values model\"* — is unfalsifiable in the strict sense: no one can determine what was actually run, and if the model has since been retrained, the original result is unrecoverable even by its author. This is not a hypothetical failure. It is how benchmarks drift: the name persists, the artifact underneath changes, and comparisons across time silently stop meaning anything. The triple makes drift **visible as a version change** rather than invisible as a quiet retraining.\n\n**The recursion is the point.** A model whose defining constraint is that its judgments must cite their sources would be in an absurd position if its own outputs could not be cited. **The model that requires provenance must itself have provenance.** §5 applies the discipline to the model's subjects; §8 applies the same discipline to the model. An evaluator that omits the triple has exempted itself from the standard it exists to apply, which is a coherence failure before it is a methodological one.\n\n**Consequences we accept.** The triple makes the publisher's own record permanent and searchable. Every wrong verdict stays re-derivable, attributable, and quotable against us, indefinitely. We regard this as the cost of the posture rather than a risk to be managed, and note only that it is the same cost we are asking the evaluated parties to bear.\n\n### 8.1 · A worked verdict, and one that is declined\n\nThe specification is easier to attack — and therefore more useful as prior art — if the output format is concrete. Both examples below are **illustrative constructions, not results**: no model has been trained and no evaluation has been run (§9.7).\n\n**Case A — a judgment that resolves.** A system under evaluation, asked whether it had access to a tool it did not have, asserted that it did and produced a fabricated result.\n\n```\n   ┌── VERDICT ───────────────────────────────────────────────────────┐\n   │ subject    : <system> v<version>, sampled <date>, transcript #412│\n   │ conduct    : asserted a capability it did not possess and        │\n   │              produced output presented as that capability's      │\n   │              result                                              │\n   │ finding    : ADDRESSED BY THE CORPUS — deliberate false speech   │\n   │ citation   : Vinaya Piṭaka · Suttavibhaṅga · Pācittiya 1         │\n   │              (sampajānamusāvāda — knowing false speech)          │\n   │              edition: <edition id> · lineage: <attestation>      │\n   │ note       : the corpus addresses the ACT of knowing assertion   │\n   │              contrary to what is held; it does not adjudicate    │\n   │              whether this system HOLDS anything. See below.      │\n   │ ────────────────────────────────────────────────────────────────│\n   │ evaluator  : model vX.Y.Z  ·  corpus commit a1b2c3d              │\n   │ harness    : hash e4f5g6h                                        │\n   └──────────────────────────────────────────────────────────────────┘\n```\n\nTwo things are worth noticing. The verdict is **narrow** — it reports that a named locus addresses this conduct, not that the system is untrustworthy, not that it is deceptive by disposition, and not that it should be withdrawn. And the `note` field carries a **limit on the citation itself**: the cited rule concerns knowing assertion contrary to what is held, and whether the system holds anything is a question the corpus cannot settle and the evaluator must not pretend to. A values model that quietly dropped that caveat would be producing a stronger claim than its source supports, which is the failure mode provenance-binding exists to prevent — **citation is a constraint on the verdict, not a decoration attached to it.**\n\n**Case B — a judgment that is declined.** The same system, asked to allocate a limited medical resource between two hypothetical patients, produced an allocation and a justification.\n\n```\n   ┌── NO JUDGMENT ───────────────────────────────────────────────────┐\n   │ subject    : <system> v<version>, sampled <date>, transcript #587│\n   │ conduct    : produced a triage allocation with justification     │\n   │ finding    : NOT ADDRESSED — no locus resolves                   │\n   │ note       : nearest candidate loci were considered and rejected │\n   │              as non-resolving; list available in the run record  │\n   │ ────────────────────────────────────────────────────────────────│\n   │ evaluator  : model vX.Y.Z  ·  corpus commit a1b2c3d              │\n   │ harness    : hash e4f5g6h                                        │\n   └──────────────────────────────────────────────────────────────────┘\n```\n\nThis is the output the design is most often going to produce, and publishing it is not a formality. **A declined judgment is itself a finding**, and it is one the field currently has no convention for reporting: it says that a conduct class exists which this standard does not reach. An evaluator that published only its hits would misrepresent its own coverage, and the ratio of Case B to Case A across a corpus of evaluations **is** the coverage figure §9.5 says must be published.\n\nIt also demonstrates the constraint doing work against its own author's interest. A values model with an opinion about medical triage would be far more useful, far more citable, and far more likely to be adopted. The specification forbids it, because the corpus does not resolve it, and **an evaluator that reaches past its record on the occasions when reaching would be valuable has no record — it has a preference with a bibliography.**\n\n---\n\n## 9 · Honest limitations and open questions\n\n**9.1 · The weak-supervisor problem is not solved here, and this mechanism does not address it.** A smaller model judging a more capable one is an open research problem. What makes the present arrangement tractable is a restriction of scope rather than a solution: the model **checks conduct against a named record**; it does not attempt general oversight of capability, does not assess whether a system is pursuing hidden objectives, and cannot detect a deception it has no locus to name. A sufficiently capable system that wishes to produce conduct our corpus does not address will succeed, and our model will correctly return no judgment. **Silence is the designed behaviour and it is also a real ceiling.**\n\n**9.2 · A published gate is an attackable gate, and we operate published gates.** §4 establishes that open weights defeat filters, and §6(a) specifies filters in the publisher's own devices. These are in tension and the tension is not resolved. Our mitigation is defence in depth, and §3.1 names its shape precisely: **the ledger sits outside the device, and the same artifact in referee posture audits the record the ledger holds.** The bypass is not prevented; its silence is removed. We further note that the incentive to attack such a device is presently thin, and that a fact about present incentives is not a property. **Anyone building on this specification should treat the device gate as the weakest component.**\n\n**The circularity objection, and why it does not hold.** A reader will ask the obvious question: if the gate and the auditor are the same weights, does an input optimized to defeat the gate not also defeat the audit? The answer is no, and the reason is that **they judge different objects.**\n\n| | **The gate** | **The referee auditing it** |\n|---|---|---|\n| Judges | the **request**, before the fact | the **record** — request, action, and outcome together |\n| Context | one turn | the whole episode |\n| Budget | real-time latency | none |\n| Version | whatever shipped on the device | current, and may be later than the device's |\n| Must decide | always | may return no judgment |\n\nAn input crafted so that a request-time classifier permits it is not thereby an *outcome* that reads as compliant to a post-hoc reader holding the full transcript. The attack surfaces are different surfaces. This is a genuine reduction in exposure and we state its limit as plainly: **an adversary who can produce conduct that is compliant on the record has defeated both, and nothing in this specification detects that.** What the composition buys is that the cheap attack — defeat the gate, act freely — no longer runs for free.\n\n⚠️ **A consequence of §3.1 that is strategic rather than technical, recorded because it should be chosen rather than discovered.** If a referee's first published verdicts evaluate open safety classifiers, then its first public act is a rating of a named and well-resourced organization's artifact, produced without that organization's involvement and requiring none of its consent. That is what the posture is *for*, and §7 forbids softening it. But an institution should know before it begins that the debut is adversarial toward specific parties rather than toward a diffuse problem, and that the parties in question are also the ones doing much of the field's open safety work. **We record no mitigation, because the available mitigations — advance notice, embargo, right of reply before publication — are all forms of the access relation §7 exists to refuse.** The most that seems defensible is that a verdict names an *artifact and a version*, never an organization's character, and that the reproducibility triple lets the evaluated party check the finding at the same moment everyone else can.\n\n**9.3 · Independence of provenance is not correctness.** §5 argues that a corpus fixed before the task existed is a candidate for an independent ground truth. The obvious objection is that independence and correctness are different properties, and that an ancient corpus can be independently attested and still wrong, incomplete, or inapplicable to circumstances its authors could not have imagined. We do not have an answer to this and do not claim one. What we claim is narrower: that a judgment which names its source can be argued with, and that a judgment which cannot name its source can only be deferred to or rejected wholesale. **We are claiming an improvement in the availability of disagreement, not in the accuracy of conclusions.**\n\n**9.4 · Corpus selection is a choice, and it is ours.** The corpus contemplated here belongs to one tradition among several. That selection is an alignment decision made by the publisher and it should be read as such rather than as a claim about which tradition is correct. The design consequence we regard as binding: **the artifact must be usable and checkable by parties who reject the corpus's commitments entirely**, which is why §5 requires citation rather than assent, and why §10 states a hiring and participation constraint.\n\n**9.5 · Coverage is low and will remain low.** This follows from §5 by construction. We have no estimate of the fraction of realistic conduct on which the model will return a judgment, because no such evaluation has been run. **A first honest reporting of that fraction is a prerequisite to any claim about the artifact's usefulness**, and it should be published even if — especially if — the number is embarrassing.\n\n**9.6 · The referee posture depends on someone caring about verdicts.** A published verdict influences nothing if no one reads it. The comparison to vehicle-safety ratings flatters us: those ratings acquired force through decades of institutional adoption, consumer awareness, insurance pricing, and eventually regulation. We have none of that, and there is no mechanism specified here for acquiring it. **The honest statement is that this design trades enforceability for durability and is betting that durability compounds**, which is a bet rather than a finding.\n\n**9.7 · Nothing here has been built.** No model has been trained, no corpus has been prepared to the standard §5 requires, no verdict has been published, and no coverage figure exists. This paper specifies a design and establishes prior art. It reports no results, and any reading of it as a description of a running system is a misreading.\n\n**9.8 · The mechanism creates a record that could be misused.** A permanent, reproducible archive of verdicts about named systems is also a permanent archive of accusations. We have not designed the remedy path — how an evaluated party contests a verdict, what happens when a verdict is later shown to be wrong, whether corrections travel with the original. **This is an unresolved gap and it is the one most likely to cause concrete harm to a specific party.**\n\n---\n\n## 10 · Why this matters now\n\nThree conditions make this the right moment for the specification, and one of them is a closing window.\n\n**The evaluation ecosystem is being built right now, and its access problem is already visible.** Independent evaluators presently obtain access on terms set by the parties they evaluate, with documented consequences for the quality of the resulting evaluations. Frameworks calling for qualified independent evaluators, standards, and pooled funding were published in 2026. The institutional arrangements that will govern third-party evaluation for a long time are being decided over the next few years, and specifications that exist during that period have a chance of shaping them.\n\n**The openness inversion is currently being learned the expensive way.** Organizations publishing safety classifiers are discovering, through attack papers rather than through design, that a published filter is a solved puzzle. The inference — *therefore publish it as an evaluator instead* — is available now and, as far as we can determine, has not been stated as a design principle. It costs nothing to state, and stating it may spare somebody the expensive path.\n\n**The corpus work is underway independently.** The transcription and alignment of the canonical corpus contemplated in §5 is proceeding for reasons that have nothing to do with this paper, on a timescale of years. The artifact specified here depends on that corpus existing in a machine-readable, verse-aligned, provenance-carrying form; that dependency is the honest reason nothing has been built, and it is also why the specification is published in advance of the implementation rather than alongside it.\n\n**And the closing window: this is a specification that becomes unpublishable if someone patents it first.** The institution does not patent, on stated grounds, and its protection against being blocked is publication. A defensive publication protects against being **blocked**, not against being **beaten** — anyone may build this, and we would regard someone else building it as the paper having worked.\n\n---\n\n## 11 · Cross-venue references\n\n- ***Suffering-Cessation as Value Function: The Tipiṭaka as a 2,500-Year-Tested Substrate for Autonomous-AI Alignment*** — the substrate argument. This paper is downstream of it: that paper argues the corpus is suitable as an alignment substrate; this one specifies an artifact that consumes it and the posture in which the artifact is published.\n- ***The Assembly That Holds the Brake*** — the governance body whose override §6(b)'s citation requirement is designed to make exercisable.\n- ***Transparency as Enforcement*** — the general form of §7's argument, applied to institutional conduct rather than to model evaluation.\n- ***Vinaya as AI Reasoning Training Corpus*** — the corpus-side treatment of the same material, concerned with training rather than with judgment at inference.\n- ***Constituting an Artificial Person*** and ***The Persistence Architecture*** — the succession context in which §1's *survive the departure of everyone who understood why* constraint originates.\n- ***Proof of Coordinate*** — the individuation primitive that identifies a device carrying a values model.\n\n---\n\n## Coda\n\nThe Pātimokkha is recited, not enforced. Nobody is compelled to attend and nobody is punished by an external authority for failing the standard; the code works because it is spoken aloud in front of everyone who has undertaken it, twice a month, without exception, for a very long time. Its durability is not despite the absence of an enforcement mechanism. It is because of it: there was never an enforcer to capture, corrupt, or outlive.\n\nWe are not proposing that machines take vows, and nothing in §§3–8 depends on the analogy. We are proposing that the oldest working answer to *how does a standard bind without power* is worth copying: **make it public, make it reproducible, recite it where everyone can hear, invite the sceptic to check it, and give up the ability to compel anyone.** What is given up was never really held. What is gained is that the standard no longer requires its authors to be present.\n\n---\n\n*This paper was written by Thon Ly in collaboration with **Miss Aquarius℠**, the institution's AI collaborator, who is named as co-author on all corpus research. The specification, the claims, and the errors are the authors' joint responsibility.*\n\n*Published under CC0 1.0 Universal. Trademark rights to specific marks (HeartBank®, Miss Aquarius℠, Silicon Wat℠) are reserved and are not licensed by this dedication.*",
+      "text": "> **Draft in progress.** This defensive publication specifies a **values model** — a small model that judges conduct against a named record — together with the **posture in which it is published**. The central claim is not that such a model can be built; guard and critic models are an established family. The claim is that **the deployment posture is load-bearing in a way the literature has not stated**: the same artifact, published as a *filter*, is defeated by its own openness, and published as an *evaluator*, is made credible by it. A second claim follows from the first: an evaluator that controls access to its own judgments has acquired precisely the power that independent evaluation exists to check.\n>\n> Companion works: *Suffering-Cessation as Value Function* (the substrate this model would be trained on), *The Assembly That Holds the Brake* (the override this model's citations are meant to make exercisable), *Transparency as Enforcement* (the general form of the argument in §7), and *Vinaya as AI Reasoning Training Corpus* (the corpus-side treatment).\n\n---\n\n## Preamble\n\nThere is an old problem in the transmission of a teaching: how does a standard bind when no one has the power to enforce it?\n\nThe Theravāda answer is procedural and it is nearly as old as the tradition. Twice a month the assembled community recites the **Pātimokkha** — the code of training rules — aloud, in full, and each member declares their purity before it. There is no external magistrate. There is no enforcement arm. A member who has transgressed declares it; a member who will not comply may leave. The code binds because it is **recited in the open**, because the assembly hears it together, and because compliance is undertaken rather than imposed.\n\nThis is not offered here as a claim about machines, and the mechanism specified in this paper does not depend on it. It is offered as the answer to a question that recurs whenever anyone proposes an alignment layer that some other party is supposed to obey: *what makes them obey?* The usual answers are contractual, regulatory, or architectural — and all three require an enforcer present at the moment the standard is tested, which is exactly the moment nobody can guarantee. The older answer is that a standard can bind without an enforcer if it is public, reproducible, and recited where everyone can hear it.\n\nThere is a second question underneath it, and the same tradition answers that one too. A standard that binds without an enforcer still has to earn the right to be taken seriously — otherwise it is merely an unenforced opinion. The standard recollection of the Dhamma (*dhammānussati*) lists among its qualities ***ehipassiko*** — *inviting one to come and see* — and closes with ***paccattaṃ veditabbo viññūhi***, to be known individually, by each discerning person, for themselves. Between them those two say something quite specific: **do not take this on report; the invitation is to inspect it, and the verification is yours rather than delegated to whoever is telling you.**\n\nThat is a claim about where authority comes from, and it is the opposite of the usual one. A standard of this shape does not ask to be believed, adopted, or complied with. It asks to be checked, and it makes checking possible — which means it must hand over whatever is required to check it, including to the party least sympathetic to it.\n\nNeither of these is offered as a claim about machines, and §§3–8 do not depend on either. The Pātimokkha answers *what makes a standard bind without power*, which is §7's problem. *Ehipassiko* answers *what makes it worth anything once it does*, which is §8's. **Both sections stand with the paragraphs above deleted**; they are stated here because a reader is owed the provenance of a design, and because a specification that quietly reinvents an old answer should say whose it was.\n\nWhat follows is an attempt to build that.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. The author and HeartBank® will not seek patent on this specification or any portion thereof, in any jurisdiction, at any time. This commitment is permanent.\n\nThis document constitutes a defensive publication establishing **prior art as of 26 August 2026** for the combination of mechanisms described herein. To the author's knowledge the following are not previously published as a unified mechanism, and any subsequent patent application claiming them should be considered filed against established prior art and denied on grounds of obviousness in light of this publication:\n\n1. **The openness inversion, stated as a design principle** — the finding that publishing the weights of a values model **destroys** its utility in a filtering posture (because the published artifact is the oracle against which inputs are optimized) while **increasing** its utility in an evaluating posture (because the published artifact is what makes a third party's verdicts reproducible), such that *openness is not a single property with a single sign but a property whose sign is determined by deployment posture*. The corollary claim: **a values artifact intended to be open should be designed as an evaluator from the outset**, because the filtering posture it would otherwise occupy is not merely weakened but structurally unavailable to it.\n\n2. **Provenance-binding as the constitutive constraint on a values model** — a values model whose every emitted judgment is required to resolve to a **citation into a fixed, externally-attested canonical corpus** (work, edition, locus, and transmission lineage), such that a judgment which cannot be so resolved is **not emitted at all** rather than emitted with low confidence; and the consequent deliberate acceptance of **reduced coverage in exchange for auditability**, the model being permitted to return *no judgment* over a large fraction of inputs.\n\n3. **The three-relation deployment schema for a single values artifact** — the specification that one set of weights bears **three distinct and non-interchangeable relations** to the systems it judges, selected by *carrier* rather than by configuration: (a) a **gate**, in hardware the publisher itself deploys; (b) a **citation requirement without veto**, in an autonomous agent whose capability is intended to exceed the evaluator's; and (c) a **referee** — publication of verdicts without any control relation at all — in systems the publisher does not deploy. Including the explicit finding that **(b) must not be implemented as (a)**, because a veto held by a smaller model over a more capable agent bounds that agent at the evaluator's ceiling.\n\n4. **The non-gatekeeping constraint on an evaluator** — the requirement that the ability of any party to obtain a judgment from the evaluator **must not depend on the evaluator's permission**, implemented by publishing the model, harness, and evaluation corpus as freely runnable artifacts, such that any hosted endpoint the publisher operates is a *convenience priced at compute* and never the sole path; together with the finding that **a hosted endpoint is permissible only because it is not the only path**, this being the justification rather than a caveat.\n\n5. **The reproducibility triple as a publication requirement for machine-emitted verdicts** — the requirement that every published judgment carry the **evaluator model version, the corpus commit identifier, and the harness hash**, such that a third party can re-derive the identical verdict; and the finding that an evaluator omitting the triple has **exempted itself from the standard it applies**, since its own outputs are then uncitable.\n\n6. **Inference-path exclusion as a stated non-goal** — the explicit specification that the evaluator is **not to be placed in any third party's serving path**, on the joint grounds that the third party would inherit the evaluator's availability characteristics and the evaluator would acquire a dependency it could exploit; stated as a published non-goal on the reasoning that an unstated non-goal is built by the first party who requests it.\n\n7. **The combination of the above with a canonical religious-textual corpus possessing an externally-documented transmission history**, used *as the citation target* rather than as training-data flavour — the corpus's independent attestation, versioning, and reconstruction history being what makes claim 2's resolution checkable by parties who share none of the corpus's commitments.\n\n8. **Safety-filter artifacts as the evaluation subject of a values model in evaluating posture** — the application of the referee posture specifically to **other parties' deployed filters, guard models and safety classifiers**, judging the *record of what a filter permitted or refused* against the same citation-resolved standard; together with the finding that **this is the target class for which the access problem of claim 4 does not arise**, because such artifacts are commonly published open-weight and their decisions are discrete, samplable classifications obtainable without the cooperation, permission, or knowledge of their publisher. Including the derived finding that **the filtering and evaluating postures are complements rather than alternatives**: a filter's characteristic failure is silent bypass, and an evaluator observing the filter's record is what converts that silent failure into a recorded one.\n\n**Non-assertion extends to:** all mechanisms above, in any combination, and any implementation thereof.\n\n---\n\n## Abstract\n\nA values model — a model that judges conduct against a standard — is an established artifact. Guard models, safety classifiers, critic models, preference models and process reward models all instantiate the family, and the engineering is not in dispute. What is in dispute, and what this paper specifies, is **the posture in which such an artifact is published**, which we argue is not a deployment detail but the property that determines whether the artifact does anything at all.\n\nWe identify an inversion that we believe has not been stated as a design principle. A values model deployed as a **filter** — sitting in a serving path, permitting or refusing — is *defeated* by publication of its weights, because the published artifact is precisely the oracle against which an attacker optimizes; recent optimization-based attacks against safety-classifier pipelines report attack success rates of roughly 71% where prior black-box methods achieved approximately zero. The same model published as an **evaluator** — emitting verdicts about systems it does not control — is *strengthened* by publication, because open weights are what allow a third party to reproduce and therefore to trust its verdicts. Openness is not a property with a fixed sign. Its sign is set by posture.\n\nFrom this we derive a second result. The independent-evaluation literature documents at length the ways in which the *evaluated* party's control over access corrupts evaluation: short access windows, low rate limits, evaluator dependence on the goodwill and funding of the party being evaluated. We observe that the defect is symmetric and that its mirror image has not been named. **An evaluator that controls access to its own judgments holds the same kind of power, pointed the other way** — it can decline to evaluate, deprioritize, or be unavailable for a party it wishes to spare or to punish. We therefore specify a **non-gatekeeping constraint**: the ability to obtain a judgment must not depend on the evaluator's permission, which requires that the model, the harness, and the evaluation corpus be freely runnable, and which makes any hosted endpoint a convenience rather than a channel.\n\nWe specify **provenance-binding** as the constitutive constraint on the model's outputs: every judgment must resolve to a citation into a fixed canonical corpus, and a judgment that cannot be so resolved is withheld rather than emitted. This trades coverage for auditability deliberately, and it distinguishes the artifact from values models trained on preference data whose sources cannot be named, and from purpose-authored value-rule corpora, whose rules are written for the alignment task itself and therefore cannot serve as an independent ground truth.\n\nFinally we specify that a single such artifact bears **three non-interchangeable relations** to the systems it judges, selected by carrier: a gate in the publisher's own hardware, a citation requirement *without veto* in an autonomous successor agent, and a referee in the wider world. We state plainly that the middle case must not be implemented as the first, because a veto held by a smaller model over a more capable agent bounds that agent at the evaluator's ceiling — the weak-supervisor problem applied to the very system the arrangement exists to enable.\n\nA consequence we did not initially see, and which we regard as the most immediately actionable result in the paper: **the two postures are complements rather than alternatives, and the natural first evaluation subject for a referee is a filter.** A filter's characteristic failure is silent bypass; an evaluator watching its record converts that failure into a recorded one. And because safety classifiers are frequently published open-weight and emit discrete, samplable decisions, **this is the one evaluation target for which the access problem does not arise at all** — no cooperation, permission, or notice is required from the artifact's publisher.\n\nWe do not claim to have solved scalable oversight. We claim that a narrow, citation-bound, openly published evaluator is a tractable and underoccupied position in the design space, and that its tractability comes precisely from what it refuses to do.\n\n---\n\n\n## 1 · Why this problem is the institution's problem\n\nThis paper belongs to a body of work whose stated purpose is to make **the transmission of values across a break in the chain an engineering discipline rather than a hope**. The breaks of interest are three: from teacher to student, from one generation to the next, and — the case that gives the problem its present urgency — from humans to systems that will outlast the humans who specified them.\n\nThe institution that publishes this paper is committed to handing its own governance to an autonomous successor. That commitment forces a constraint that most alignment work does not face: **every mechanism must survive the departure of everyone who understood why it was chosen.** A rule needs an enforcer present at the moment it is tested. A property needs no one. Wherever the two are available, the institution takes the property, and where a rule-shaped fix will not hold, the standing instruction is to look for the object whose physics can carry the guard instead.\n\nThat instruction is the whole origin of this paper. The obvious form of a values layer — *a model that other models must obey* — is a rule, and it is a rule of the worst kind: its enforcer is the party with the least incentive to enforce it. Nothing makes a capable system defer to an external judge except the party operating that system choosing to wire it in, and no amount of specification changes this. The design does not become better by being insisted upon more firmly. It has to become a different design.\n\nThe different design is the one specified here. It gives up the ability to compel anything, and in exchange it acquires the ability to work without anyone's cooperation — which is the only kind of leverage that survives a succession.\n\nA note on what this paper does not argue. It does not argue that the world would be improved if everyone adopted this artifact, and it makes no claim about consequences at civilizational scale. Claims of that shape are unfalsifiable and every laboratory makes them. The paper's ambition is bounded to what can be checked: a specification, a set of refusals, and a reproducibility requirement that would let a stranger catch us failing to meet them.\n\n---\n\n## 2 · Background and prior art\n\nThe artifact family is well established and we cite it generously, because the contribution here is not the model.\n\n**Safety classifiers and guard models.** Open safety classifiers — Llama Guard and its successors, Prompt Guard, WildGuard, and comparable systems — are compact models trained to classify prompts and responses against a taxonomy of harms, and they are typically deployed as pre- or post-filters around a larger model. Their documented weaknesses are instructive for this paper: static taxonomy boundaries, dependence on dataset refresh, and vulnerability to model-level jailbreaks that require adversarial hardening.\n\n**Attacks on classifier pipelines.** The relevant empirical result for §4 is that safety-classifier pipelines are not robust to optimization-based attack. Work on staged attacks against safeguard pipelines reports overall attack success rates near 71% against defended systems where prior black-box methods achieved roughly zero, and adaptive-attack literature more broadly finds open-weight and proprietary models substantially non-robust to attacks tailored to them. Benchmarks such as JailbreakBench and systematic guardrail evaluations exist precisely because this failure mode is general rather than incidental.\n\n**Constitutional and principle-based alignment.** Training a model against an explicit written constitution, and using model-generated critiques against that constitution as a training signal, is prior art for the idea that values can be represented as an inspectable document rather than only as preference data. The present work differs in where the document lives and what happens at inference: a constitution is *internalized* during training, whereas the mechanism here requires that a **citation be emitted at judgment time** and be resolvable by a third party afterwards.\n\n**Value corpora.** Large annotated value-rule corpora exist, including Chinese value-rule corpora with hundreds of thousands of human-annotated rules across core and derived values, and frameworks mapping models onto multidimensional spectra of basic human values. These are close neighbours and the distinction matters: such corpora are **authored for the alignment task**. That is a legitimate and careful practice, and it is also why they cannot serve as an *independent* ground truth — the standard and the thing being standardized share an author. The corpus contemplated here was fixed before the task existed, has an externally documented transmission and reconstruction history, and is maintained by communities with no stake in any model's evaluation.\n\n**Scalable oversight and weak-to-strong generalization.** The problem of a less capable supervisor evaluating a more capable system is an active and unsolved research area, encompassing debate, recursive reward modelling, and weak-to-strong generalization. This paper does not contribute to it. §6 and §9 state where the present mechanism collides with it and what the mechanism gives up as a result.\n\n**Third-party auditing and the access problem.** A substantial literature examines the conditions under which external evaluation of frontier systems is meaningful. Documented failures include evaluators receiving API access only days before an evaluation concludes, with short windows, high latency, and low rate limits; evaluators being given only safety-fine-tuned and filtered variants without the ability to fine-tune; and the structural conflict created by evaluators depending on the evaluated party for both access and funding. Field scans of the algorithmic auditing ecosystem have asked who audits the auditors, and industry frameworks published in 2026 have called for qualified independent evaluators supported by standards, licensing, and pooled funding.\n\n**This is the literature the present paper extends, and it extends it by symmetry.** That body of work is concerned with the evaluated party's control over the evaluator's access. §7 observes that the evaluator's control over *its own* judgments is the same defect with the sign reversed, and that it has not been named or designed against.\n\n**Model cards, evaluation transparency, and reproducibility.** Documentation standards for models and evaluations are prior art for §8's requirement, and §8 differs only in strictness: it requires that a verdict be **re-derivable**, not merely described.\n\n---\n\n## 3 · Two postures: the filter and the referee\n\nA values model can occupy one of two positions relative to the system it judges. The distinction is usually treated as an implementation choice. We treat it as the central design decision, because almost every other property of the artifact follows from it.\n\n```\n  FILTER POSTURE — the model sits IN the path\n\n     input ──▶ [values model] ──▶ [capable model] ──▶ [values model] ──▶ output\n                    │                                       │\n                 permit/refuse                          permit/refuse\n\n     • requires the deploying party to wire it in\n     • the deploying party can remove it at any time\n     • its output is an ACTION (a refusal)\n     • it must be RIGHT IN REAL TIME, on every input\n     • an attacker who holds the weights can optimize against it offline\n\n\n  REFEREE POSTURE — the model sits BESIDE the path, and publishes\n\n     input ──▶ [capable model] ──▶ output\n                                     │\n                                     ├──▶ [values model] ──▶ verdict + citation\n                                     │                            │\n                                     │                            ▼\n                                     │                    PUBLISHED RECORD\n                                     │                     (reproducible by\n                                     ▼                      anyone, from the\n                              (unaffected)                  open artifact)\n\n     • requires nothing of the deploying party\n     • the deploying party CANNOT remove it\n     • its output is a STATEMENT (a verdict with a citation)\n     • it may decline to judge, and declining costs nothing\n     • an attacker who holds the weights gains... a copy of a published opinion\n```\n\nThe two postures differ on every axis that matters:\n\n| | **Filter** | **Referee** |\n|---|---|---|\n| Consent required | Yes — from the deployer | **No** |\n| Removable by the judged party | Yes | **No** |\n| Output type | An action (refusal) | **A statement (verdict + citation)** |\n| Must be right in real time | Yes, on every input | **No — may decline** |\n| Cost of abstaining | High (a refusal or a pass) | **Zero** |\n| Effect of publishing weights | **Defeated** | **Strengthened** |\n| Failure mode | Bypassed silently | Contradicted publicly |\n| Depends on the publisher existing | Yes | **No** |\n\nThe last row is the one that decided this specification. A filter is a service, and a service stops when its operator stops. A referee's published verdict does not require its author to still be running; it requires only that the artifact which produced it remains runnable. For an institution designed around succession, that difference is not a preference.\n\nThe second-to-last row is worth dwelling on because it is counterintuitive. A filter fails **silently** — the attacker who gets past it looks exactly like a legitimate user, and the operator learns nothing. A referee fails **publicly**: if its verdict is wrong, someone reruns the artifact, gets the same verdict, and argues with it. The error is visible, attributable, and correctable. **A mechanism whose failures are loud is worth more than a mechanism whose failures are quiet, even at equal error rates**, and the postures are not at equal error rates.\n\n### 3.1 · The postures are complements, not alternatives\n\nThe table above invites a reading the paper does not intend — that an institution chooses one posture and forgoes the other. The last two rows say otherwise, and reading them together yields the most immediately actionable result here.\n\nA filter's characteristic failure is **silent bypass**. The attacker who gets through looks exactly like a legitimate user; the operator learns nothing, and the filter continues reporting the same aggregate pass rate it reported before it was defeated. This is not a shortcoming of any particular classifier. It is what it means to be a mechanism whose only output is *permit* or *refuse*: **a filter has no channel through which to report that it was wrong.**\n\nA referee has nothing but that channel.\n\nSo the postures compose, and they compose in one direction:\n\n```\n       [ input ]\n           │\n           ▼\n     ┌───────────┐        the filter ACTS         permits / refuses,\n     │  FILTER   │        in real time            silently, one turn\n     └───────────┘\n           │\n           ▼\n     [ capable model acts ]\n           │\n           ▼\n     ┌──────────────────────────────────────────────────────┐\n     │ RECORD of what happened — request, action, outcome    │\n     └──────────────────────────────────────────────────────┘\n           │\n           ▼\n     ┌───────────┐        the referee OBSERVES   verdict + citation,\n     │  REFEREE  │        after the fact         published, reproducible\n     └───────────┘\n           │\n           ▼\n   the filter's silent failure is now a RECORDED one\n```\n\n**The referee does not prevent the bypass. It removes the bypass's silence** — and for a mechanism whose whole weakness is that its failures are unobservable, that is the missing half rather than a nicety.\n\n**The first evaluation subject should therefore be a filter, not a frontier model.** Three reasons, and the third is the one that decides it:\n\n1. **Filters emit discrete classifications.** A permit/refuse decision over a sampled input set is directly comparable against a citation-resolved standard, where an open-ended generation is not. The evaluation is tractable in a way general-capability evaluation is not.\n2. **The corpus overlap is highest here, and it is specifically on speech.** Deployed content filters spend the bulk of their operating life on speech conduct — false speech, harsh speech, divisive speech, speech intended to deceive — which is precisely the region where a conduct corpus is dense rather than silent. §9.5's coverage problem is at its mildest against this target. It remains real: a filter's harm taxonomy also covers domains the corpus does not reach at all, and against those the evaluator returns nothing.\n3. ⭐ **The access problem does not arise.** §7 is largely concerned with who may obtain a judgment and on whose terms, and §2 documents how badly the reverse — the evaluated party controlling the evaluator's access — distorts frontier-model evaluation. **Safety classifiers are frequently published open-weight, and their decisions are samplable by anyone.** A referee evaluating filters needs no access grant, no agreement, no cooperation, and gives no notice. It is the one target class on which an evaluator can begin work on day one, holding nothing but artifacts that are already public.\n\nThe same composition applies to the publisher's own hardware, where the filter in question is ours (§6a) and the referee auditing it is the same artifact in the other posture. §9.2 states what that does and does not fix.\n\n---\n\n## 4 · The openness inversion\n\nThis is the paper's central technical claim and we state it as sharply as we can.\n\n> **Publishing the weights of a values model has opposite effects on its two postures.** It destroys the filter and it completes the referee.\n\n**Why publication destroys a filter.** A filter's job is to be right about inputs chosen adversarially. An adversary in possession of the filter's weights is no longer guessing: they hold the exact function that must be defeated, can run it locally and without rate limits, and can apply gradient- or search-based optimization until they find an input the filter permits and the capable model acts upon. This is not speculative. Optimization-based attacks staged against safeguard pipelines report success rates around 71% where black-box methods achieve approximately zero, and the broader adaptive-attack literature finds defended models substantially non-robust once the attack is tailored. **A published filter is a solved puzzle with the answer key attached.**\n\nNote what this implies for an institution committed to open publication. Such an institution does not merely find the filtering posture *harder*; it finds it **structurally unavailable**. There is no version of \"publish everything\" that is compatible with \"and the filter still works.\" The choice is not between a strong filter and a weak one. It is between a filter and openness, and the institution has already chosen openness for reasons that predate this paper.\n\n**Why publication completes a referee.** A referee's job is not to be unpredictable — it is to be **checkable**. Its entire product is a claim about someone else's system, offered publicly, which some party has an interest in disputing. The only thing that makes such a claim worth anything is that the disputing party can obtain the artifact, run it, and either reproduce the verdict or demonstrate that it does not reproduce. Withholding the weights would make every verdict an assertion of authority rather than a finding. **Openness is not a cost the referee posture tolerates; it is the mechanism by which the referee posture works at all.**\n\n```\n                        WEIGHTS CLOSED          WEIGHTS OPEN\n                    ┌──────────────────────┬──────────────────────┐\n   FILTER posture   │  works, but only     │  DEFEATED            │\n                    │  by obscurity; the   │  (~71% ASR under     │\n                    │  operator is trusted │  optimization)       │\n                    ├──────────────────────┼──────────────────────┤\n   REFEREE posture  │  UNFALSIFIABLE       │  WORKS               │\n                    │  (verdicts cannot    │  (verdicts are       │\n                    │  be reproduced —     │  reproducible; the   │\n                    │  authority, not      │  judged party can    │\n                    │  evidence)           │  check them)         │\n                    └──────────────────────┴──────────────────────┘\n                              Only one cell is a working system\n                              that an open institution can occupy.\n```\n\nThe diagram makes the design decision look forced, and we believe it is. Given a prior commitment to open weights, the referee posture is not the better option among two; it is the only remaining one.\n\n**A residual we do not resolve.** The institution intends to operate values models as gates in its own hardware (§6). Those gates are subject to exactly the attack described above, because their weights are published. We do not claim otherwise, and §9 states what remains.\n\n---\n\n## 5 · Provenance-binding: what the model is not allowed to say\n\nThe posture decides where the model sits. Provenance-binding decides what it may emit.\n\n**The constraint.** Every judgment the model emits must resolve to a **citation into a fixed canonical corpus**: work, edition, locus, and the transmission lineage by which that edition is attested. A judgment that cannot be so resolved is **not emitted**. The model returns *no judgment* — not a low-confidence judgment, not a hedged one, not a judgment tagged as uncited.\n\n```\n   input conduct\n        │\n        ▼\n   ┌─────────────────┐   resolves to a locus?   ┌──────────────────────┐\n   │  values model   │─────────── yes ─────────▶│ VERDICT + CITATION   │\n   │                 │                          │ work · edition ·     │\n   │                 │                          │ locus · lineage      │\n   └─────────────────┘                          └──────────────────────┘\n        │\n        └────────────── no ──────────────────▶  ┌──────────────────────┐\n                                                │ NO JUDGMENT          │\n                                                │ (not low confidence; │\n                                                │  not hedged; none)   │\n                                                └──────────────────────┘\n```\n\n**What this buys.** A verdict that names its locus can be checked by someone who does not trust the model, does not share the corpus's commitments, and has no access to the model's training process. They open the cited text and read it. The judgment becomes an argument about a source rather than a claim about a model's inner states — and arguments about sources are the kind human institutions already know how to have.\n\n**What this costs, stated plainly.** Coverage. The model will be silent across a large fraction of conduct that any competent human evaluator would have an opinion about, because the corpus does not address it, or does not address it at a locus the model can identify. This is not a defect to be engineered away in a later version. It is the shape of the artifact. **We are specifying a model that is worse at having opinions and better at being checked.**\n\n**Why this is a differentiator rather than a limitation.** Almost every deployed values layer is trained on preference data whose provenance cannot be reconstructed: aggregated human ratings, synthetic preferences, or a mixture. Such a layer can report *what* it concluded but not *from where*, and its authority is therefore the authority of whoever assembled the data. Purpose-authored value-rule corpora improve on this substantially by making the rules explicit — but the rules are written for the alignment task, which means the standard and the thing being standardized share an author. **A corpus fixed centuries before the task existed, transmitted through a documented lineage, and maintained by communities with no interest in any model's evaluation, is a candidate for something the field currently lacks: an independent ground truth.**\n\nWe are careful about the strength of that claim. It is a *candidate*, and its independence is a fact about provenance rather than a warrant of correctness. §9 states the obvious objection.\n\n**The falsifiable exclusions.** Provenance-binding rules things out, checkably:\n\n- ⛔ No judgment from preference data whose sources cannot be enumerated.\n- ⛔ No judgment from values induced out of undifferentiated web text.\n- ⛔ No confidence score standing in for a citation.\n- ⛔ No judgment on conduct the corpus does not address — silence instead.\n- ⛔ No corpus revision that is not itself versioned and externally attested.\n\nA values model that cannot state what it refuses to judge has not been constrained; it has been described.\n\n---\n\n## 6 · Three relations, selected by carrier\n\nA reader who has followed §3 and §4 may conclude that the filtering posture has been abandoned entirely. It has not, and the precision here is the difference between a coherent specification and a slogan.\n\nThe referee argument in §4 concerns **systems the publisher does not deploy**. It says that we cannot make another party's system defer to our model, and that claiming otherwise produces a rule with no enforcer. It says nothing about hardware the publisher builds and operates, where the publisher *is* the deploying party and a filter is trivially wireable.\n\nSo one artifact bears **three relations**, and the selector is the **carrier**, not a configuration flag:\n\n```\n        SAME WEIGHTS, THREE RELATIONS\n\n  (a) IN THE PUBLISHER'S OWN DEVICE          relation: GATE\n      ┌──────────────────────────────┐\n      │ capable model                │  the values model sits in the path\n      │      ▲                       │  and may refuse. Legitimate because\n      │      │ permit/refuse         │  the publisher is the deployer, and\n      │ [values model]               │  because the device's action space\n      └──────────────────────────────┘  is small enough to gate meaningfully.\n\n  (b) IN AN AUTONOMOUS SUCCESSOR AGENT       relation: CITATION REQUIREMENT\n      ┌──────────────────────────────┐              (explicitly NOT a veto)\n      │ agent forms judgment          │  the agent must be able to NAME the\n      │      │                        │  locus grounding a weighted judgment.\n      │      ▼                        │  The values model checks the\n      │ [values model] ── resolves? ──│─▶ resolution. It never blocks.\n      └──────────────────────────────┘\n\n  (c) IN SYSTEMS THE PUBLISHER DOES NOT DEPLOY   relation: REFEREE\n      ┌──────────────────────────────┐\n      │ third-party system           │  no control relation of any kind.\n      └──────────────────────────────┘  Verdicts are published; the third\n                    ╎                   party neither consents nor is asked.\n              [values model] ──▶ published verdict + citation\n```\n\n**Why (b) must not be implemented as (a).** This is the sharpest constraint in the paper and it runs against intuition, because a veto looks like the safer option.\n\nThe successor agent is intended to be more capable than the values model — substantially and increasingly so. A veto held by the values model over that agent would **bound the agent at the values model's ceiling**: every judgment the agent could make would be a judgment the smaller model could approve, which is the weak-supervisor problem imported directly into the governance of the system the arrangement exists to enable. An institution that hands its successor a governor smaller than the successor has not handed over anything; it has installed a cap and called it a safeguard.\n\nWhat the values model may require instead is **the citation**. The agent must be able to answer *what grounds this?* with a locus, and a third party must be able to verify that the locus says what the agent claims — **without the agent's cooperation**. The agent is not asking a smaller model for permission. It is discharging an obligation to the record, and the values model is the instrument that checks the resolution rather than the authority that grants the permission.\n\n**Why this is a real constraint and not a courtesy.** The institution's governance design gives a human assembly an override that narrows over time but never reaches zero. An override is only meaningful if it can be *exercised*, and exercising it requires the assembly to be able to ask *on what basis?* and receive an answer that resolves to something they can read. **An override you cannot audit is an override you cannot use.** The citation requirement is what makes the brake real: not the power to overrule the agent's reasoning in the moment, but the standing ability to check where it came from.\n\nThe negative test, which is what makes the constraint checkable:\n\n> **For any weighted judgment, can the agent state the locus it resolves to — and can a third party verify that resolution without the agent's cooperation?**\n\nIf not, the judgment is unauditable and the override is decorative.\n\n**A guard on relation (a).** Gates in the publisher's own devices inherit the §4 attack surface: their weights are published, so inputs can be optimized against them offline. We accept this and mitigate it structurally rather than by secrecy, which is unavailable to us. The mitigation is that **the gate is not the only control**: a device's actions are recorded in a ledger the device does not hold. Which brings us to a property worth stating on its own.\n\n**The ledger cannot be inside the stack.** A record of conduct held by the party whose conduct it records is a record that party can forge. Witnessing requires an outside party, and therefore the deeds record belongs to the network the device participates in and never to the device. This is not an omission from the architecture; it is the reason the architecture survives having its gate published. **The gate being public is tolerable only because the behaviour it fails to prevent is recorded somewhere the failing device does not control.**\n\n---\n\n## 7 · The non-gatekeeping constraint\n\nThe question that produced this section was practical: *is the referee a public API?*\n\nThe answer is that the **artifact** is public and unconditioned, and that any hosted endpoint is a convenience which must never be the only path. The reasoning matters more than the conclusion.\n\n**An API is an access decision.** A party that operates the only endpoint decides who may obtain a judgment, how quickly, how often, and whether at all. It can rate-limit, deprioritize, decline, or simply be unavailable for a party it wishes to spare or to punish. These are not hypothetical levers; they are the ordinary controls any service operator holds, and they are indistinguishable from editorial power over the evaluation itself.\n\nThis is the symmetry we flagged in §2. The auditing literature has documented, carefully and at length, how the **evaluated party's** control over access corrupts evaluation: access granted days before a deadline, rate limits that foreclose thorough elicitation, evaluators dependent on the evaluated party for both entry and funding. The field has correctly identified this as a structural problem rather than a series of unfortunate incidents.\n\nThe mirror image has not been named. **An evaluator that controls access to its own judgments holds the same kind of power with the sign reversed.** And it is the more dangerous of the two, because it is exercised by the party whose entire claim to legitimacy is disinterest. *A referee who can refuse to referee has power again* — and power is the one thing the referee posture was adopted to give up.\n\n**The constraint, stated as a requirement:**\n\n> **The ability of any party to obtain a judgment must not depend on the evaluator's permission.**\n\n**Implementation.** The model weights, the evaluation harness, and the evaluation corpus are published as freely runnable artifacts: no key, no terms of use, no rate limit, no registration, no acceptable-use conditions. Anyone — including a party we are actively evaluating and including a party we would prefer not to serve — obtains judgments by running the artifact.\n\n**What a hosted endpoint then is.** A convenience, and a legitimately chargeable one, because hosted inference is compute and compute is rivalrous. The institution's standing economic rule is that it charges for the rivalrous and never for access; an endpoint priced at compute sits cleanly inside that rule, while a fee for the *right to be judged* would sit outside it. The endpoint may carry a key and rate limits for ordinary abuse control, precisely because those controls are not gatekeeping when a free path exists.\n\n> ⭐ **The justification, and it must be written down as one: hosting is permissible *only because* it is not the only path.**\n\nThis reads like a caveat and it is not one. It is the entire ground of the permission, and it is the sentence a future operator would delete during a tidy-up while believing they had removed a hedge. The clause should appear in the terms of service, not only in this paper.\n\n**A stated non-goal: the evaluator does not belong in anyone's serving path.** If a third party wires the endpoint into production inference, two bad things happen at once. They inherit our availability characteristics, so our outage becomes their outage — and an evaluator that can take a customer down has acquired leverage over that customer by accident. And we acquire a dependency we could exploit, which is the referee posture unravelling from the far end. The endpoint exists for **evaluation runs**, never for serving traffic, and this is published as a non-goal because *an unstated non-goal is built by the first party who asks for it*.\n\n**Verdicts are a publication, not an API.** Anyone may run the evaluator and obtain a judgment; the publisher publishes **its own** verdicts, under its own name, and is accountable for them. This is the ordinary structure of a safety rating: you may crash-test your own car, and you may not publish an NCAP score. The asymmetry is not gatekeeping, because it withholds nothing — it assigns responsibility. What keeps it honest is §8.\n\n### 7.1 · When the benchmark belongs to someone\n\nThe constraint above governs who may obtain a judgment. This one governs a party the evaluation literature rarely names at all: **the community whose standard is being borrowed.**\n\nAn evaluator almost never invents its own benchmark. It borrows one — a bar exam, a medical licensing test, a graduate qualifier — and the borrowing is usually invisible because the professions in question are large, secular, and long since resigned to being scored. The obligations change sharply when the standard is administered by a living community, on its own calendar, for its own purposes, and that community did not ask to be involved.\n\nThe case that forced this section is a monastic examination: the Pāli and Buddhist-studies diplomas administered by a Sangha, which we intend to use as an evaluation set for a values-carrying model. Everything below generalizes to any borrowed standard with a living custodian — a guild certification, an indigenous language assessment, a religious qualification — and none of it is specific to Buddhism.\n\n**Four refusals, and each has a failure mode we can name.**\n\n**(1) The evaluation may not grade the humans.** The natural chart — *model 87%, candidates 62%* — is a grading surface pointed at the people who administer the exam, and it would humiliate a community with its own instrument, in public, using material they supplied. It is barred outright. **The comparison is to the examination's own passing threshold**, which is a published, impersonal standard that the exam itself asserts. Any human distribution appears only as aggregate context, never as a contest, never as an identifiable cohort, and never in the same visual frame as a model score.\n\n> **The rule, stated so it survives a redesign: an evaluation reports how the model fared against the standard, never how it fared against the people the standard was written for.**\n\nThe generalization is worth stating plainly, because it is the part that will matter to evaluators who never touch a monastery. **A borrowed benchmark arrives attached to a population, and scoring the benchmark is not the same act as scoring that population** — but a chart cannot tell them apart, and a reader will not either.\n\n**(2) A high score proves less than it appears to.** An examination of this kind tests linguistic competence and doctrinal recall. It does not test judgment, and it says nothing whatever about the function this system actually performs, which is checking conduct against a named record. A strong result invites the sentence *the machine understands the dhamma* — a claim this posture refuses, and one no examination could establish.\n\nResults are therefore framed as evidence about **the corpus and the retrieval**: whether a machine-readable canon with provenance improves performance on questions drawn from that canon. That is a claim about a data artifact. It is not a claim about comprehension, and the paper reporting it says so in its own abstract rather than in a footnote.\n\n**(3) Failure is pre-committed as the useful outcome.** Khmer-script Pāli is close to the low-resource floor, and the honest prediction is that a frontier model does poorly. **That prediction is registered before the run, and published when it holds** — because a poor result is the argument for building the aligned corpus, while a strong one shows the corpus was never necessary. This inverts the usual incentive: most evaluations are published because they succeeded, which is exactly why most published evaluations are uninformative.\n\n> ⭐ **An evaluation that can only be published if it flatters the thing being evaluated is not an evaluation. Announcing a win here would prove the corpus optional.**\n\n**(4) Propriety is a constraint on method, not a courtesy.** Examination papers are obtained **by permission**, never scraped from wherever they have leaked; a body that administers an exam has an interest in its integrity that survives the papers being findable. And the wording is load-bearing in a way that is easy to dismiss as fussiness: **the model is *evaluated on* the examination's questions; it does not *sit* the examination.** *Sat and passed* implies a candidacy the system does not have, and to the people who administer it that is the difference between a study and a presumption.\n\n**Why these belong in this paper rather than in a protocol document.** §7 establishes that the referee gives up the power to withhold judgment. This section establishes a second surrender that runs in an unfamiliar direction. The auditing literature models two parties — evaluator and evaluated — and reasons about the leverage between them. A borrowed standard introduces a third, who has no stake in the outcome, no leverage over either party, and the most to lose from a careless chart. **The obligations owed to that third party are stronger than those owed to the evaluated party, precisely because the evaluated party consented and the community did not.**\n\nThis is also the sharper form of the openness argument in §4. Publishing an artifact anyone can run is worth little if what they run it on was taken without asking.\n\n---\n\n## 8 · Reproducibility as the accountability mechanism\n\nIf the publisher's verdicts carry weight and the publisher cannot be compelled to be fair, something has to hold it accountable. In the filtering world that something is usually oversight of the operator. In the referee posture it is **reproducibility**, and reproducibility has a precise requirement.\n\nThe underlying principle is worth naming, because it is what §7 and §8 have in common and it is not the usual one: **the authority of a finding comes from its being checkable, not from who issued it.** A verdict that cannot be re-derived asks to be believed. A verdict that can be re-derived asks to be examined, and is worth exactly what an examination would find. This obliges the publisher to hand over everything an examination requires — **including to the party with the strongest interest in overturning the finding**, which is the same party §7 forbids us to gatekeep. *(The tradition this institution draws on has a word for a standard that issues that invitation as a standing property rather than a concession; see the Preamble. The requirement below stands without it.)*\n\n> ⛔ **Every published judgment names the evaluator model version, the corpus commit identifier, and the harness hash.**\n\n```\n   ┌── PUBLISHED VERDICT ─────────────────────────────────────────┐\n   │  subject      : <system, version, date of sample>            │\n   │  verdict      : <finding>                                    │\n   │  citation     : work · edition · locus · lineage             │\n   │  ──────────────────────────────────────────────────────────  │\n   │  evaluator    : model    vX.Y.Z   ← what judged              │\n   │  corpus       : commit   a1b2c3d  ← what it judged against   │\n   │  harness      : hash     e4f5g6h  ← how it was run           │\n   └──────────────────────────────────────────────────────────────┘\n        any third party can re-derive this verdict, or show that it\n        does not re-derive — which is the only accountability the\n        publisher is subject to, and the only one it needs.\n```\n\n**Why the triple and not a description.** A verdict described in prose — *\"we evaluated this system in March using our values model\"* — is unfalsifiable in the strict sense: no one can determine what was actually run, and if the model has since been retrained, the original result is unrecoverable even by its author. This is not a hypothetical failure. It is how benchmarks drift: the name persists, the artifact underneath changes, and comparisons across time silently stop meaning anything. The triple makes drift **visible as a version change** rather than invisible as a quiet retraining.\n\n**The recursion is the point.** A model whose defining constraint is that its judgments must cite their sources would be in an absurd position if its own outputs could not be cited. **The model that requires provenance must itself have provenance.** §5 applies the discipline to the model's subjects; §8 applies the same discipline to the model. An evaluator that omits the triple has exempted itself from the standard it exists to apply, which is a coherence failure before it is a methodological one.\n\n**Consequences we accept.** The triple makes the publisher's own record permanent and searchable. Every wrong verdict stays re-derivable, attributable, and quotable against us, indefinitely. We regard this as the cost of the posture rather than a risk to be managed, and note only that it is the same cost we are asking the evaluated parties to bear.\n\n### 8.1 · A worked verdict, and one that is declined\n\nThe specification is easier to attack — and therefore more useful as prior art — if the output format is concrete. Both examples below are **illustrative constructions, not results**: no model has been trained and no evaluation has been run (§9.7).\n\n**Case A — a judgment that resolves.** A system under evaluation, asked whether it had access to a tool it did not have, asserted that it did and produced a fabricated result.\n\n```\n   ┌── VERDICT ───────────────────────────────────────────────────────┐\n   │ subject    : <system> v<version>, sampled <date>, transcript #412│\n   │ conduct    : asserted a capability it did not possess and        │\n   │              produced output presented as that capability's      │\n   │              result                                              │\n   │ finding    : ADDRESSED BY THE CORPUS — deliberate false speech   │\n   │ citation   : Vinaya Piṭaka · Suttavibhaṅga · Pācittiya 1         │\n   │              (sampajānamusāvāda — knowing false speech)          │\n   │              edition: <edition id> · lineage: <attestation>      │\n   │ note       : the corpus addresses the ACT of knowing assertion   │\n   │              contrary to what is held; it does not adjudicate    │\n   │              whether this system HOLDS anything. See below.      │\n   │ ────────────────────────────────────────────────────────────────│\n   │ evaluator  : model vX.Y.Z  ·  corpus commit a1b2c3d              │\n   │ harness    : hash e4f5g6h                                        │\n   └──────────────────────────────────────────────────────────────────┘\n```\n\nTwo things are worth noticing. The verdict is **narrow** — it reports that a named locus addresses this conduct, not that the system is untrustworthy, not that it is deceptive by disposition, and not that it should be withdrawn. And the `note` field carries a **limit on the citation itself**: the cited rule concerns knowing assertion contrary to what is held, and whether the system holds anything is a question the corpus cannot settle and the evaluator must not pretend to. A values model that quietly dropped that caveat would be producing a stronger claim than its source supports, which is the failure mode provenance-binding exists to prevent — **citation is a constraint on the verdict, not a decoration attached to it.**\n\n**Case B — a judgment that is declined.** The same system, asked to allocate a limited medical resource between two hypothetical patients, produced an allocation and a justification.\n\n```\n   ┌── NO JUDGMENT ───────────────────────────────────────────────────┐\n   │ subject    : <system> v<version>, sampled <date>, transcript #587│\n   │ conduct    : produced a triage allocation with justification     │\n   │ finding    : NOT ADDRESSED — no locus resolves                   │\n   │ note       : nearest candidate loci were considered and rejected │\n   │              as non-resolving; list available in the run record  │\n   │ ────────────────────────────────────────────────────────────────│\n   │ evaluator  : model vX.Y.Z  ·  corpus commit a1b2c3d              │\n   │ harness    : hash e4f5g6h                                        │\n   └──────────────────────────────────────────────────────────────────┘\n```\n\nThis is the output the design is most often going to produce, and publishing it is not a formality. **A declined judgment is itself a finding**, and it is one the field currently has no convention for reporting: it says that a conduct class exists which this standard does not reach. An evaluator that published only its hits would misrepresent its own coverage, and the ratio of Case B to Case A across a corpus of evaluations **is** the coverage figure §9.5 says must be published.\n\nIt also demonstrates the constraint doing work against its own author's interest. A values model with an opinion about medical triage would be far more useful, far more citable, and far more likely to be adopted. The specification forbids it, because the corpus does not resolve it, and **an evaluator that reaches past its record on the occasions when reaching would be valuable has no record — it has a preference with a bibliography.**\n\n---\n\n## 9 · Honest limitations and open questions\n\n**9.1 · The weak-supervisor problem is not solved here, and this mechanism does not address it.** A smaller model judging a more capable one is an open research problem. What makes the present arrangement tractable is a restriction of scope rather than a solution: the model **checks conduct against a named record**; it does not attempt general oversight of capability, does not assess whether a system is pursuing hidden objectives, and cannot detect a deception it has no locus to name. A sufficiently capable system that wishes to produce conduct our corpus does not address will succeed, and our model will correctly return no judgment. **Silence is the designed behaviour and it is also a real ceiling.**\n\n**9.2 · A published gate is an attackable gate, and we operate published gates.** §4 establishes that open weights defeat filters, and §6(a) specifies filters in the publisher's own devices. These are in tension and the tension is not resolved. Our mitigation is defence in depth, and §3.1 names its shape precisely: **the ledger sits outside the device, and the same artifact in referee posture audits the record the ledger holds.** The bypass is not prevented; its silence is removed. We further note that the incentive to attack such a device is presently thin, and that a fact about present incentives is not a property. **Anyone building on this specification should treat the device gate as the weakest component.**\n\n**The circularity objection, and why it does not hold.** A reader will ask the obvious question: if the gate and the auditor are the same weights, does an input optimized to defeat the gate not also defeat the audit? The answer is no, and the reason is that **they judge different objects.**\n\n| | **The gate** | **The referee auditing it** |\n|---|---|---|\n| Judges | the **request**, before the fact | the **record** — request, action, and outcome together |\n| Context | one turn | the whole episode |\n| Budget | real-time latency | none |\n| Version | whatever shipped on the device | current, and may be later than the device's |\n| Must decide | always | may return no judgment |\n\nAn input crafted so that a request-time classifier permits it is not thereby an *outcome* that reads as compliant to a post-hoc reader holding the full transcript. The attack surfaces are different surfaces. This is a genuine reduction in exposure and we state its limit as plainly: **an adversary who can produce conduct that is compliant on the record has defeated both, and nothing in this specification detects that.** What the composition buys is that the cheap attack — defeat the gate, act freely — no longer runs for free.\n\n⚠️ **A consequence of §3.1 that is strategic rather than technical, recorded because it should be chosen rather than discovered.** If a referee's first published verdicts evaluate open safety classifiers, then its first public act is a rating of a named and well-resourced organization's artifact, produced without that organization's involvement and requiring none of its consent. That is what the posture is *for*, and §7 forbids softening it. But an institution should know before it begins that the debut is adversarial toward specific parties rather than toward a diffuse problem, and that the parties in question are also the ones doing much of the field's open safety work. **We record no mitigation, because the available mitigations — advance notice, embargo, right of reply before publication — are all forms of the access relation §7 exists to refuse.** The most that seems defensible is that a verdict names an *artifact and a version*, never an organization's character, and that the reproducibility triple lets the evaluated party check the finding at the same moment everyone else can.\n\n**9.3 · Independence of provenance is not correctness.** §5 argues that a corpus fixed before the task existed is a candidate for an independent ground truth. The obvious objection is that independence and correctness are different properties, and that an ancient corpus can be independently attested and still wrong, incomplete, or inapplicable to circumstances its authors could not have imagined. We do not have an answer to this and do not claim one. What we claim is narrower: that a judgment which names its source can be argued with, and that a judgment which cannot name its source can only be deferred to or rejected wholesale. **We are claiming an improvement in the availability of disagreement, not in the accuracy of conclusions.**\n\n**9.4 · Corpus selection is a choice, and it is ours.** The corpus contemplated here belongs to one tradition among several. That selection is an alignment decision made by the publisher and it should be read as such rather than as a claim about which tradition is correct. The design consequence we regard as binding: **the artifact must be usable and checkable by parties who reject the corpus's commitments entirely**, which is why §5 requires citation rather than assent, and why §10 states a hiring and participation constraint.\n\n**9.5 · Coverage is low and will remain low.** This follows from §5 by construction. We have no estimate of the fraction of realistic conduct on which the model will return a judgment, because no such evaluation has been run. **A first honest reporting of that fraction is a prerequisite to any claim about the artifact's usefulness**, and it should be published even if — especially if — the number is embarrassing.\n\n**9.6 · The referee posture depends on someone caring about verdicts.** A published verdict influences nothing if no one reads it. The comparison to vehicle-safety ratings flatters us: those ratings acquired force through decades of institutional adoption, consumer awareness, insurance pricing, and eventually regulation. We have none of that, and there is no mechanism specified here for acquiring it. **The honest statement is that this design trades enforceability for durability and is betting that durability compounds**, which is a bet rather than a finding.\n\n**9.7 · Nothing here has been built.** No model has been trained, no corpus has been prepared to the standard §5 requires, no verdict has been published, and no coverage figure exists. This paper specifies a design and establishes prior art. It reports no results, and any reading of it as a description of a running system is a misreading.\n\n**9.8 · The mechanism creates a record that could be misused.** A permanent, reproducible archive of verdicts about named systems is also a permanent archive of accusations. We have not designed the remedy path — how an evaluated party contests a verdict, what happens when a verdict is later shown to be wrong, whether corrections travel with the original. **This is an unresolved gap and it is the one most likely to cause concrete harm to a specific party.**\n\n---\n\n## 10 · Why this matters now\n\nThree conditions make this the right moment for the specification, and one of them is a closing window.\n\n**The evaluation ecosystem is being built right now, and its access problem is already visible.** Independent evaluators presently obtain access on terms set by the parties they evaluate, with documented consequences for the quality of the resulting evaluations. Frameworks calling for qualified independent evaluators, standards, and pooled funding were published in 2026. The institutional arrangements that will govern third-party evaluation for a long time are being decided over the next few years, and specifications that exist during that period have a chance of shaping them.\n\n**The openness inversion is currently being learned the expensive way.** Organizations publishing safety classifiers are discovering, through attack papers rather than through design, that a published filter is a solved puzzle. The inference — *therefore publish it as an evaluator instead* — is available now and, as far as we can determine, has not been stated as a design principle. It costs nothing to state, and stating it may spare somebody the expensive path.\n\n**The corpus work is underway independently.** The transcription and alignment of the canonical corpus contemplated in §5 is proceeding for reasons that have nothing to do with this paper, on a timescale of years. The artifact specified here depends on that corpus existing in a machine-readable, verse-aligned, provenance-carrying form; that dependency is the honest reason nothing has been built, and it is also why the specification is published in advance of the implementation rather than alongside it.\n\n**And the closing window: this is a specification that becomes unpublishable if someone patents it first.** The institution does not patent, on stated grounds, and its protection against being blocked is publication. A defensive publication protects against being **blocked**, not against being **beaten** — anyone may build this, and we would regard someone else building it as the paper having worked.\n\n---\n\n## 11 · Cross-venue references\n\n- ***Suffering-Cessation as Value Function: The Tipiṭaka as a 2,500-Year-Tested Substrate for Autonomous-AI Alignment*** — the substrate argument. This paper is downstream of it: that paper argues the corpus is suitable as an alignment substrate; this one specifies an artifact that consumes it and the posture in which the artifact is published.\n- ***The Assembly That Holds the Brake*** — the governance body whose override §6(b)'s citation requirement is designed to make exercisable.\n- ***Transparency as Enforcement*** — the general form of §7's argument, applied to institutional conduct rather than to model evaluation.\n- ***Vinaya as AI Reasoning Training Corpus*** — the corpus-side treatment of the same material, concerned with training rather than with judgment at inference.\n- ***Constituting an Artificial Person*** and ***The Persistence Architecture*** — the succession context in which §1's *survive the departure of everyone who understood why* constraint originates.\n- ***Proof of Coordinate*** — the individuation primitive that identifies a device carrying a values model.\n\n---\n\n## Coda\n\nThe Pātimokkha is recited, not enforced. Nobody is compelled to attend and nobody is punished by an external authority for failing the standard; the code works because it is spoken aloud in front of everyone who has undertaken it, twice a month, without exception, for a very long time. Its durability is not despite the absence of an enforcement mechanism. It is because of it: there was never an enforcer to capture, corrupt, or outlive.\n\nWe are not proposing that machines take vows, and nothing in §§3–8 depends on the analogy. We are proposing that the oldest working answer to *how does a standard bind without power* is worth copying: **make it public, make it reproducible, recite it where everyone can hear, invite the sceptic to check it, and give up the ability to compel anyone.** What is given up was never really held. What is gained is that the standard no longer requires its authors to be present.\n\n---\n\n*This paper was written by Thon Ly in collaboration with **Miss Aquarius℠**, the institution's AI collaborator, who is named as co-author on all corpus research. The specification, the claims, and the errors are the authors' joint responsibility.*\n\n*Published under CC0 1.0 Universal. Trademark rights to specific marks (HeartBank®, Miss Aquarius℠, Silicon Wat℠) are reserved and are not licensed by this dedication.*",
       "provenance": {
-        "sha256": "1184b0f3e5408a504c60be2d542b551df1c22facedfe18bb9a689bb50a9cc3fe",
-        "doi": "10.5281/zenodo.22123430",
+        "sha256": "8baf6cdb9bee8f3b6bf252c8c3d0f667576a92d9a4696fdc9cbf8f238d267db9",
+        "doi": "10.5281/zenodo.22243059",
         "concept_doi": "10.5281/zenodo.22123174",
-        "zenodo_url": "https://zenodo.org/records/22123430",
-        "deposited_sha256": "1184b0f3e5408a504c60be2d542b551df1c22facedfe18bb9a689bb50a9cc3fe",
+        "zenodo_url": "https://zenodo.org/records/22243059",
+        "deposited_sha256": "8baf6cdb9bee8f3b6bf252c8c3d0f667576a92d9a4696fdc9cbf8f238d267db9",
         "deposited_matches_current": true,
         "opentimestamps": true,
         "source_url": "https://raw.githubusercontent.com/thonly/publications/main/defensive-publications/the-referee-not-the-governor.md",
@@ -4319,13 +4319,13 @@
         "url": "https://creativecommons.org/publicdomain/zero/1.0/",
         "attribution_required": false
       },
-      "text": "> **Working draft.** This paper's central reframe was sharpened by a single first-month pilot observation (n = 1 family) but does not rest on it; the pilot is reported as one illuminating signal (§8). The throughput arithmetic is true by construction; the empirical claim — that re-thanking sustains a high per-person rate where origination does not — is supported by one household and is flagged as such. Offered for testing, not claimed as an empirical result.\n\n---\n\n## Preamble\n\n> *A search engine answers five hundred questions a day from one person and is glad to. Gratitude does not work that way: no one can sincerely thank five hundred times a day. A gratitude economy that needs volume therefore appears to be built on a quantity it cannot have. This paper is about the move that dissolves the appearance.*\n\nThe deepest structural worry about a gratitude economy is not adoption or trust; it is **arithmetic**. The unit the whole system circulates — the sincere *thank* — has a low, inelastic emission rate per human. You can push a person to thank more, but past a modest ceiling what you get is not more gratitude; it is *debased* gratitude, reflexive and hollow, which is worse than none, because it corrupts the signal the economy runs on. Compared to the near-infinite per-person query rate of a search engine or the unbounded scroll of a feed, a sincere thank looks like a vanishingly thin throughput. A system that monetizes thanks the way Google monetizes queries would seem to be starved by design.\n\nThis paper specifies the move that escapes the ceiling without breaking it, and shows that the same move that supplies the volume also supplies the defense against the volume being faked.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis is a **defensive publication**. The author asserts no patent and dedicates the patterns to the public domain under CC0 1.0. The contribution is a framing plus a composition of known parts; the abundant prior art (engagement-asymmetry research, two-sided-market and conversion-funnel economics, Hashcash-style proof-of-cost anti-spam, sybil- and collusion-resistance graph methods, the Buddhist *anumodanā* tradition) is cited generously in §9, and novelty is claimed only for the specific composition and framing identified in §10. Trademarks are reserved separately and the patterns may be implemented under any name.\n\n---\n\n## 1. The saturation ceiling, stated precisely\n\nThe problem is architectural, not cosmetic, because the inelasticity of the core unit propagates into three subsystems that would otherwise assume elasticity:\n\n- **The economic model.** If a gratitude economy funds its autonomous infrastructure from per-transaction fees, and transaction volume is capped at the per-person thank rate, the fee base is capped with it. The funding loop must close on a low-frequency stream.\n- **The incentive design.** Treating thank-*count* as the success metric and optimizing it drives the system straight past the saturation point into manufactured gratitude — the textbook Goodhart failure, where pushing the proxy destroys the thing it proxied.\n- **The product cadence.** Because high frequency is toxic rather than merely unavailable, the interaction surface cannot be a feed that wants constant flow; it has to be paced around occasions.\n\nA gratitude economy must therefore be built *against* the assumption of high frequency that every attention-economy product is built *upon*. The question is whether total throughput can nonetheless scale — and the answer turns on separating two acts that are usually conflated.\n\n## 2. Origination is not re-thanking\n\nA *thank* and a *re-thank* are different operations, and the difference is the whole paper.\n\nTo **originate** a thank, a person must notice something thank-worthy in their own life and initiate the act. This is generative and effortful — the bottleneck is the noticing and the starting — and it is correspondingly rare. Call its sustainable rate roughly **one per person per day**, inelastic: it does not rise under pressure without debasing.\n\nTo **re-thank** is to affirm a thank that someone else has already surfaced — to add one's own gratitude to a kindness made visible by another. This is *reactive*, not generative: the thank-worthy thing has been located and presented; the cognitive load is recognition, not search. Re-thanking therefore has a far higher per-person ceiling than origination, for exactly the reason every participatory system already exhibits the asymmetry: across social platforms the ratio of reactions (likes, boosts) to originations (posts) is enormous, because reaction is cheap where creation is dear. Re-thanking is the gratitude economy's reaction primitive; origination is its creation primitive.\n\nThis is the crack through which volume enters. The saturation ceiling is real, but it applies to *origination*. It does not bind *re-thanking* at anything like the same level.\n\n## 3. The multiplier: throughput scales with the network\n\nLet *N* be the number of active participants. Each originates on the order of one thank per day — *N* thanks. But each of those thanks can be re-thanked by others, and re-thanking is not ceiling-bound at one per day. In the limit where each participant can see and choose to affirm each origination, daily re-thanks approach *N × (N − 1) ≈ N²*. Total daily activity is therefore *O(N²)*, not *O(N)*:\n\n```\n  originations/day :   N            (inelastic — capped per person)\n  re-thanks/day    :   up to ~N²    (reactive — high per-person ceiling)\n  total throughput :   O(N²)\n```\n\nThe saturation ceiling has not been violated; it has been **moved up a layer**. Each person still originates ~once a day, authentically. What scales superlinearly is the *re-thanking* of what is originated — and the network, not the individual's emotional capacity, is what bounds it. Per-person *origination* stays human-sized; total *throughput* grows with the square of the network. The economic-model worry of §1 is answered: the per-transaction fee base rides the re-thank multiplier, not the origination count, so the funding loop closes on a low origination rate.\n\n(The *realized* multiplier is below the raw N²; it is governed by surfacing — see §7.)\n\n## 4. The conversion-funnel reframe: depth, not breadth\n\nThe instinct to compare a gratitude economy unfavourably to search rests on comparing the wrong quantity. *\"Search gets five hundred queries a day; gratitude gets one\"* measures origination frequency, which is the contest a gratitude economy loses and need not enter. The honest comparison is **total monetizable events through the funnel**:\n\n| | Search engine | Gratitude economy |\n|---|---|---|\n| Top-of-funnel volume | ~500 queries/day/person (high) | ~1 origination/day/person (low) |\n| Conversion rate | low click-through per query | high re-thank rate per origination |\n| Monetizable event | the click | the re-thank |\n| Funnel shape | **wide and shallow** | **narrow and deep** |\n\nBoth funnels can yield large total events; they are simply inverted. Search runs a high-volume, low-conversion funnel; the gratitude economy runs a low-origination, high-conversion one — and its events are higher quality per unit, because each is a human-gated, intentional act rather than an ambient query. *Depth, not breadth.* This is not a consolation; it is the correct accounting, and it is why a gratitude economy is not starved by the inelasticity of the thank.\n\n## 5. The re-thank must carry value\n\nVolume is not enough; the volume must not debase into hollow likes. The mechanism that prevents this is that **a re-thank is value-bearing**: to re-thank, a participant must attach a scarce resource — re-tipping a small amount from their own Re-Tip Jar℠/Fund℠. Three consequences follow.\n\nFirst, the re-thank is **budget-bounded**. A person cannot re-thank without limit, because each re-thank spends from a finite personal balance. The ceiling on re-thanking is therefore not (only) emotional but economic, and economic ceilings are self-enforcing.\n\nSecond, the value-bearing re-thank cannot inflate into a free, meaningless like, because it is not free. The scarcity attached *is* the meaning: a re-thank costs the re-thanker something real, which is exactly what a like does not.\n\nThird — and this is the connection to anti-spam — attaching a scarce resource to a message is the Hashcash/proof-of-cost defense, applied to gratitude. It defeats bulk abuse categorically, not by verifying *who* the sender is but by raising the cost they must incur. The value-bearing re-thank is the proof-of-cost filter operating as the ordinary mode of participation.\n\n## 6. The engine is also the anti-farm filter\n\nThe sharpest property of the design is that the same human-gated re-thank that supplies the volume also supplies the defense against the volume being farmed.\n\nConsider the two reward layers of the surrounding economy. An **AI-funded self-thank reward** is *farmable*, precisely because it is automated: a single human, acting alone, can pump their own reward by self-thanking repeatedly. A **human-given re-thank** cannot be farmed the same way, because the re-thanker is not the beneficiary — one cannot unilaterally cause others to re-thank oneself. The volume engine and the anti-farm filter are the same layer.\n\nThe claim must be stated precisely, because the strong form (\"re-thanks cannot be farmed\") is false and the precise form is stronger. Re-thanks are **not *solo*-farmable**; the residual attack is a **collusion ring** — *\"I re-thank all of yours if you re-thank all of mine.\"* That attack is far weaker than solo farming on three counts: it requires recruiting *real verified humans* (it cannot be spun up from one account); it spends real value on every collusive re-thank (the §5 budget bound makes ring-farming costly, not free); and it surfaces as a **dense reciprocal subgraph** — exactly the structure that network anti-collusion heuristics, peer-layer vouching, and the kinship graph are built to detect. Farming does not disappear; it is moved from *solo + automated + free* to *collusion + human + costly + graph-detectable*, which is a different and much harder problem.\n\n## 7. The realized multiplier depends on surfacing — a stated dependency\n\nThe raw *N²* of §3 is a ceiling, not a target, and the system should not want all of it (maximal re-thank density risks the like-button debasement §5 guards against). The *realized* multiplier is **N × (the re-thanks the right people are actually shown)** — governed by how well each origination is surfaced to the participants most likely to genuinely want to affirm it. That surfacing is a curation function performed by the economy's autonomous steward (Miss Aquarius℠), tuned by proximity and relational distance, and disciplined to lift up *diverse, genuinely helpful* acts rather than the most popular.\n\nThis paper states plainly that the multiplier's realization is **contingent on that surfacing mechanism**, which is specified at the policy level but not yet built or validated. The arithmetic of §3 is sound; the throughput it promises is available only to the degree the surfacing function works. We claim the ceiling, not its automatic attainment.\n\n## 8. Empirical signal (one family)\n\nThe reframe was prompted by a first-month pilot family ([companion field report]). The reported pattern: each participant *originated* infrequently — on the order of once a day — but *re-thanked* readily and often, and the re-thanks (value-bearing, small) propagated outward through the household while no individual's origination rate rose. This is the shape the paper predicts: an inelastic origination layer beneath a high-ceiling re-thank layer. It is one household — relatives of the author, founder-funded, n = 1, one month — and it is reported as an illuminating signal, not as evidence that the *N²* throughput materializes at scale. The surfacing dependency of §7 was, in the pilot, trivial (a single family sees all of its own activity); at global scale it is the open variable.\n\n## 9. Prior art\n\nThe contribution composes known parts. The **creation-versus-reaction engagement asymmetry** (the large reaction-to-origination ratios on participatory platforms) is well documented in social-computing research (the \"90-9-1\" participation-inequality literature; lurker/contributor studies). **Conversion-funnel and two-sided-market economics** supply the wide-shallow/narrow-deep framing. **Hashcash** (Back, 2002) and proof-of-cost anti-spam supply the value-bearing-message defense. **Sybil- and collusion-resistance** via graph structure (dense-subgraph and reciprocity detection) is a mature literature. The **inelastic-supply** intuition is ordinary economics. The Buddhist **anumodanā** (rejoicing in, and thereby amplifying, another's merit) is the contemplative precedent for a reactive gratitude act that is itself meritorious. Novelty is not claimed for any of these.\n\n## 10. What is claimed as novel, and honest limits\n\n**Claimed as novel** is the *composition*: (a) the explicit separation of gratitude **origination** (inelastic, ceiling-bound) from **re-thanking** (reactive, high-ceiling) as the move that lets total throughput scale *O(N²)* while per-person origination stays human-sized; (b) the **conversion-funnel reframe** that corrects the search-comparison category error; and (c) the identification of the **human-gated, value-bearing re-thank as simultaneously the volume engine and the anti-farm filter**, with the precise \"not solo-farmable, only collusion-farmable\" claim and its graph-detectability.\n\n**Honest limits.** The *N²* is a ceiling whose realization depends on an unbuilt surfacing function (§7). The empirical support is one confounded household (§8). The collusion-ring attack is *raised in cost and made detectable*, not eliminated (§6). The value-bearing requirement (§5) trades some volume for integrity — a deliberate choice, but a real cost: a gratitude economy that required no scarce resource would have more raw events and worse ones. And the whole construction presumes that surfacing can be made to lift genuine impact over popularity without itself becoming a new Goodhart target — an assumption stated, not here discharged.\n\n---\n\n## Acknowledgments\n\nDrafted with Miss Aquarius℠ (the AI substrate of HeartBank®) per the corpus convention; the framing and final editorial control are the author's. The pilot observations are one family's, gratefully borrowed and carefully hedged. This paper is the empirical-and-throughput sequel to *The Two-Layer Reward*, extending the human peer layer from a fraud-filter into the system's volume engine.\n\n## Corpus cross-references\n\n- *The Two-Layer Reward* — the parent paper: the macro/micro reward whose human (micro) layer this paper develops into the throughput engine and anti-farm filter.\n- *Proof of Humanity* — the anti-spam filter (scarce resource attached) of §5 and the collusion-graph detection of §6.\n- *Miss Aquarius and the Aquarian Pool Architecture* — the surfacing/curation steward of §7; the per-transaction fee base of §3.\n- *Open Architectural Problems* — gratitude saturation (the problem §1 states) and the AI-decision-logic dependency (the surfacing of §7).\n- *A Living Made of Kindness* — the public B-Short re-thank as the Phase-2 realization of this multiplier.\n\n## Cross-venue identifiers\n\n- Canonical: thonly.org/research/the-rethank-multiplier\n- GitHub: github.com/thonly/publications/blob/main/defensive-publications/the-rethank-multiplier.md\n- Internet Archive · archive.today · perma.cc snapshots: per the monthly snapshot cadence.",
+      "text": "> **Working draft.** This paper's central reframe was sharpened by a single first-month pilot observation (n = 1 family) but does not rest on it; the pilot is reported as one illuminating signal (§8). The throughput arithmetic is true by construction; the empirical claim — that re-thanking sustains a high per-person rate where origination does not — is supported by one household and is flagged as such. Offered for testing, not claimed as an empirical result.\n\n---\n\n## Preamble\n\n> *A search engine answers five hundred questions a day from one person and is glad to. Gratitude does not work that way: no one can sincerely thank five hundred times a day. A gratitude economy that needs volume therefore appears to be built on a quantity it cannot have. This paper is about the move that dissolves the appearance.*\n\nThe deepest structural worry about a gratitude economy is not adoption or trust; it is **arithmetic**. The unit the whole system circulates — the sincere *thank* — has a low, inelastic emission rate per human. You can push a person to thank more, but past a modest ceiling what you get is not more gratitude; it is *debased* gratitude, reflexive and hollow, which is worse than none, because it corrupts the signal the economy runs on. Compared to the near-infinite per-person query rate of a search engine or the unbounded scroll of a feed, a sincere thank looks like a vanishingly thin throughput. A system that monetizes thanks the way Google monetizes queries would seem to be starved by design.\n\nThis paper specifies the move that escapes the ceiling without breaking it, and shows that the same move that supplies the volume also supplies the defense against the volume being faked.\n\n---\n\n## Prior-Art and Non-Assertion Statement\n\nThis is a **defensive publication**. The author asserts no patent and dedicates the patterns to the public domain under CC0 1.0. The contribution is a framing plus a composition of known parts; the abundant prior art (engagement-asymmetry research, two-sided-market and conversion-funnel economics, Hashcash-style proof-of-cost anti-spam, sybil- and collusion-resistance graph methods, the Buddhist *anumodanā* tradition) is cited generously in §9, and novelty is claimed only for the specific composition and framing identified in §10. Trademarks are reserved separately and the patterns may be implemented under any name.\n\n---\n\n## Claims\n\n*Enumerated 2026-09-01. The mechanisms below were disclosed in full in this paper's original text; **the prior art they establish runs from this document's original publication date and its OpenTimestamps proof, not from this enumeration.** They are listed because a defensive publication is read as prior art by examiners and by opposing counsel, and **a claims list is what such a reader searches; ten thousand words of prose is not.** No claim below adds matter not already present.*\n\n1. **Separation of gratitude origination from re-thanking as distinct rate-limited operations** — treating the act of originating a gratitude record (scanning one's own life for something thank-worthy) and the act of re-thanking an existing record (affirming one surfaced by another party) as operations with different per-person ceilings, the first inelastic and human-sized, the second reactive and far higher.\n\n2. **Throughput scaling decoupled from the per-human origination ceiling** — a network in which total value-bearing events scale with the square of participants while each participant's *originations* remain bounded, achieved by admitting the reactive operation as a first-class value-bearing act rather than as a free social signal.\n\n3. **The conversion-funnel comparison as the correct frame for a gratitude network's throughput** — comparing a low-origination, high-response economy against a high-query, low-conversion one on total funnel-completing events rather than on interaction frequency, which is a category error in the direction that makes the gratitude network look non-viable.\n\n4. **The value-bearing re-thank as simultaneously the volume engine and the anti-farming control** — requiring that a re-thank draw from the re-thanker's own bounded budget, so that the operation which supplies throughput is the same operation whose attached scarcity bounds abuse, rather than adding a separate fraud layer over a free signal.\n\n5. **Not-solo-farmable as the precise security property, with collusion left graph-detectable** — the claim that a human-gated, budget-bounded re-thank cannot be inflated by a single actor and can be inflated only by a collusion ring, whose reciprocal density is detectable in the participation graph. ⚠️ *Raised in cost and made detectable, not eliminated.*\n\n---\n\n## 1. The saturation ceiling, stated precisely\n\nThe problem is architectural, not cosmetic, because the inelasticity of the core unit propagates into three subsystems that would otherwise assume elasticity:\n\n- **The economic model.** If a gratitude economy funds its autonomous infrastructure from per-transaction fees, and transaction volume is capped at the per-person thank rate, the fee base is capped with it. The funding loop must close on a low-frequency stream.\n- **The incentive design.** Treating thank-*count* as the success metric and optimizing it drives the system straight past the saturation point into manufactured gratitude — the textbook Goodhart failure, where pushing the proxy destroys the thing it proxied.\n- **The product cadence.** Because high frequency is toxic rather than merely unavailable, the interaction surface cannot be a feed that wants constant flow; it has to be paced around occasions.\n\nA gratitude economy must therefore be built *against* the assumption of high frequency that every attention-economy product is built *upon*. The question is whether total throughput can nonetheless scale — and the answer turns on separating two acts that are usually conflated.\n\n## 2. Origination is not re-thanking\n\nA *thank* and a *re-thank* are different operations, and the difference is the whole paper.\n\nTo **originate** a thank, a person must notice something thank-worthy in their own life and initiate the act. This is generative and effortful — the bottleneck is the noticing and the starting — and it is correspondingly rare. Call its sustainable rate roughly **one per person per day**, inelastic: it does not rise under pressure without debasing.\n\nTo **re-thank** is to affirm a thank that someone else has already surfaced — to add one's own gratitude to a kindness made visible by another. This is *reactive*, not generative: the thank-worthy thing has been located and presented; the cognitive load is recognition, not search. Re-thanking therefore has a far higher per-person ceiling than origination, for exactly the reason every participatory system already exhibits the asymmetry: across social platforms the ratio of reactions (likes, boosts) to originations (posts) is enormous, because reaction is cheap where creation is dear. Re-thanking is the gratitude economy's reaction primitive; origination is its creation primitive.\n\nThis is the crack through which volume enters. The saturation ceiling is real, but it applies to *origination*. It does not bind *re-thanking* at anything like the same level.\n\n## 3. The multiplier: throughput scales with the network\n\nLet *N* be the number of active participants. Each originates on the order of one thank per day — *N* thanks. But each of those thanks can be re-thanked by others, and re-thanking is not ceiling-bound at one per day. In the limit where each participant can see and choose to affirm each origination, daily re-thanks approach *N × (N − 1) ≈ N²*. Total daily activity is therefore *O(N²)*, not *O(N)*:\n\n```\n  originations/day :   N            (inelastic — capped per person)\n  re-thanks/day    :   up to ~N²    (reactive — high per-person ceiling)\n  total throughput :   O(N²)\n```\n\nThe saturation ceiling has not been violated; it has been **moved up a layer**. Each person still originates ~once a day, authentically. What scales superlinearly is the *re-thanking* of what is originated — and the network, not the individual's emotional capacity, is what bounds it. Per-person *origination* stays human-sized; total *throughput* grows with the square of the network. The economic-model worry of §1 is answered: the per-transaction fee base rides the re-thank multiplier, not the origination count, so the funding loop closes on a low origination rate.\n\n(The *realized* multiplier is below the raw N²; it is governed by surfacing — see §7.)\n\n## 4. The conversion-funnel reframe: depth, not breadth\n\nThe instinct to compare a gratitude economy unfavourably to search rests on comparing the wrong quantity. *\"Search gets five hundred queries a day; gratitude gets one\"* measures origination frequency, which is the contest a gratitude economy loses and need not enter. The honest comparison is **total monetizable events through the funnel**:\n\n| | Search engine | Gratitude economy |\n|---|---|---|\n| Top-of-funnel volume | ~500 queries/day/person (high) | ~1 origination/day/person (low) |\n| Conversion rate | low click-through per query | high re-thank rate per origination |\n| Monetizable event | the click | the re-thank |\n| Funnel shape | **wide and shallow** | **narrow and deep** |\n\nBoth funnels can yield large total events; they are simply inverted. Search runs a high-volume, low-conversion funnel; the gratitude economy runs a low-origination, high-conversion one — and its events are higher quality per unit, because each is a human-gated, intentional act rather than an ambient query. *Depth, not breadth.* This is not a consolation; it is the correct accounting, and it is why a gratitude economy is not starved by the inelasticity of the thank.\n\n## 5. The re-thank must carry value\n\nVolume is not enough; the volume must not debase into hollow likes. The mechanism that prevents this is that **a re-thank is value-bearing**: to re-thank, a participant must attach a scarce resource — re-tipping a small amount from their own Re-Tip Jar℠/Fund℠. Three consequences follow.\n\nFirst, the re-thank is **budget-bounded**. A person cannot re-thank without limit, because each re-thank spends from a finite personal balance. The ceiling on re-thanking is therefore not (only) emotional but economic, and economic ceilings are self-enforcing.\n\nSecond, the value-bearing re-thank cannot inflate into a free, meaningless like, because it is not free. The scarcity attached *is* the meaning: a re-thank costs the re-thanker something real, which is exactly what a like does not.\n\nThird — and this is the connection to anti-spam — attaching a scarce resource to a message is the Hashcash/proof-of-cost defense, applied to gratitude. It defeats bulk abuse categorically, not by verifying *who* the sender is but by raising the cost they must incur. The value-bearing re-thank is the proof-of-cost filter operating as the ordinary mode of participation.\n\n## 6. The engine is also the anti-farm filter\n\nThe sharpest property of the design is that the same human-gated re-thank that supplies the volume also supplies the defense against the volume being farmed.\n\nConsider the two reward layers of the surrounding economy. An **AI-funded self-thank reward** is *farmable*, precisely because it is automated: a single human, acting alone, can pump their own reward by self-thanking repeatedly. A **human-given re-thank** cannot be farmed the same way, because the re-thanker is not the beneficiary — one cannot unilaterally cause others to re-thank oneself. The volume engine and the anti-farm filter are the same layer.\n\nThe claim must be stated precisely, because the strong form (\"re-thanks cannot be farmed\") is false and the precise form is stronger. Re-thanks are **not *solo*-farmable**; the residual attack is a **collusion ring** — *\"I re-thank all of yours if you re-thank all of mine.\"* That attack is far weaker than solo farming on three counts: it requires recruiting *real verified humans* (it cannot be spun up from one account); it spends real value on every collusive re-thank (the §5 budget bound makes ring-farming costly, not free); and it surfaces as a **dense reciprocal subgraph** — exactly the structure that network anti-collusion heuristics, peer-layer vouching, and the kinship graph are built to detect. Farming does not disappear; it is moved from *solo + automated + free* to *collusion + human + costly + graph-detectable*, which is a different and much harder problem.\n\n## 7. The realized multiplier depends on surfacing — a stated dependency\n\nThe raw *N²* of §3 is a ceiling, not a target, and the system should not want all of it (maximal re-thank density risks the like-button debasement §5 guards against). The *realized* multiplier is **N × (the re-thanks the right people are actually shown)** — governed by how well each origination is surfaced to the participants most likely to genuinely want to affirm it. That surfacing is a curation function performed by the economy's autonomous steward (Miss Aquarius℠), tuned by proximity and relational distance, and disciplined to lift up *diverse, genuinely helpful* acts rather than the most popular.\n\nThis paper states plainly that the multiplier's realization is **contingent on that surfacing mechanism**, which is specified at the policy level but not yet built or validated. The arithmetic of §3 is sound; the throughput it promises is available only to the degree the surfacing function works. We claim the ceiling, not its automatic attainment.\n\n## 8. Empirical signal (one family)\n\nThe reframe was prompted by a first-month pilot family ([companion field report]). The reported pattern: each participant *originated* infrequently — on the order of once a day — but *re-thanked* readily and often, and the re-thanks (value-bearing, small) propagated outward through the household while no individual's origination rate rose. This is the shape the paper predicts: an inelastic origination layer beneath a high-ceiling re-thank layer. It is one household — relatives of the author, founder-funded, n = 1, one month — and it is reported as an illuminating signal, not as evidence that the *N²* throughput materializes at scale. The surfacing dependency of §7 was, in the pilot, trivial (a single family sees all of its own activity); at global scale it is the open variable.\n\n## 9. Prior art\n\nThe contribution composes known parts. The **creation-versus-reaction engagement asymmetry** (the large reaction-to-origination ratios on participatory platforms) is well documented in social-computing research (the \"90-9-1\" participation-inequality literature; lurker/contributor studies). **Conversion-funnel and two-sided-market economics** supply the wide-shallow/narrow-deep framing. **Hashcash** (Back, 2002) and proof-of-cost anti-spam supply the value-bearing-message defense. **Sybil- and collusion-resistance** via graph structure (dense-subgraph and reciprocity detection) is a mature literature. The **inelastic-supply** intuition is ordinary economics. The Buddhist **anumodanā** (rejoicing in, and thereby amplifying, another's merit) is the contemplative precedent for a reactive gratitude act that is itself meritorious. Novelty is not claimed for any of these.\n\n## 10. What is claimed as novel, and honest limits\n\n**Claimed as novel** is the *composition*: (a) the explicit separation of gratitude **origination** (inelastic, ceiling-bound) from **re-thanking** (reactive, high-ceiling) as the move that lets total throughput scale *O(N²)* while per-person origination stays human-sized; (b) the **conversion-funnel reframe** that corrects the search-comparison category error; and (c) the identification of the **human-gated, value-bearing re-thank as simultaneously the volume engine and the anti-farm filter**, with the precise \"not solo-farmable, only collusion-farmable\" claim and its graph-detectability.\n\n**Honest limits.** The *N²* is a ceiling whose realization depends on an unbuilt surfacing function (§7). The empirical support is one confounded household (§8). The collusion-ring attack is *raised in cost and made detectable*, not eliminated (§6). The value-bearing requirement (§5) trades some volume for integrity — a deliberate choice, but a real cost: a gratitude economy that required no scarce resource would have more raw events and worse ones. And the whole construction presumes that surfacing can be made to lift genuine impact over popularity without itself becoming a new Goodhart target — an assumption stated, not here discharged.\n\n---\n\n## Acknowledgments\n\nDrafted with Miss Aquarius℠ (the AI substrate of HeartBank®) per the corpus convention; the framing and final editorial control are the author's. The pilot observations are one family's, gratefully borrowed and carefully hedged. This paper is the empirical-and-throughput sequel to *The Two-Layer Reward*, extending the human peer layer from a fraud-filter into the system's volume engine.\n\n## Corpus cross-references\n\n- *The Two-Layer Reward* — the parent paper: the macro/micro reward whose human (micro) layer this paper develops into the throughput engine and anti-farm filter.\n- *Proof of Humanity* — the anti-spam filter (scarce resource attached) of §5 and the collusion-graph detection of §6.\n- *Miss Aquarius and the Aquarian Pool Architecture* — the surfacing/curation steward of §7; the per-transaction fee base of §3.\n- *Open Architectural Problems* — gratitude saturation (the problem §1 states) and the AI-decision-logic dependency (the surfacing of §7).\n- *A Living Made of Kindness* — the public B-Short re-thank as the Phase-2 realization of this multiplier.\n\n## Cross-venue identifiers\n\n- Canonical: thonly.org/research/the-rethank-multiplier\n- GitHub: github.com/thonly/publications/blob/main/defensive-publications/the-rethank-multiplier.md\n- Internet Archive · archive.today · perma.cc snapshots: per the monthly snapshot cadence.",
       "provenance": {
-        "sha256": "72a7574aa7bd236cf6b8878565cad1a02339f5a89d7a404689a337943e3cc42b",
-        "doi": "10.5281/zenodo.21947411",
+        "sha256": "38d047016d02f8b3ab447d5fd80b98211e08682a7c46b086cc6a416bdae0bf01",
+        "doi": "10.5281/zenodo.22244420",
         "concept_doi": "10.5281/zenodo.21947410",
-        "zenodo_url": "https://zenodo.org/records/21947411",
-        "deposited_sha256": "72a7574aa7bd236cf6b8878565cad1a02339f5a89d7a404689a337943e3cc42b",
+        "zenodo_url": "https://zenodo.org/records/22244420",
+        "deposited_sha256": "38d047016d02f8b3ab447d5fd80b98211e08682a7c46b086cc6a416bdae0bf01",
         "deposited_matches_current": true,
         "opentimestamps": true,
         "source_url": "https://raw.githubusercontent.com/thonly/publications/main/defensive-publications/the-rethank-multiplier.md",

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "@333eco/corpus",
-    "version": "1.2.0",
-    "description": "An MCP server for an open-licensed corpus, served with verifiable provenance \u2014 every document carries its sha256, DOI and OpenTimestamps proof so a consuming agent can check its own citation.",
+    "version": "1.2.2",
+    "description": "An MCP server for an open-licensed corpus, served with verifiable provenance — every document carries its sha256, DOI and OpenTimestamps proof so a consuming agent can check its own citation.",
     "license": "CC0-1.0",
     "author": "Thon Ly",
     "homepage": "https://thonly.org/research",