@lapxo/obligations 0.4.5 → 0.4.6

package/contracts/extension-lineage-participation.json

@@ -0,0 +1,30 @@
+{
+  "boundary": "Definitions settle meanings only. Different hypotheses or missing proofs are not a fork. No algebra, actor identity rule, M/3-M/4, protocol or wire changes.",
+  "classification": "definitions and conditional interpretation; no new law",
+  "definitions": {
+    "effectiveParticipation": "For finite nonnegative masses a_i with positive total, w_i=a_i/sum(a), K_eff=1/sum(w_i^2). Dimensionless; zeros do not count as active support. It measures weight concentration, not causally independent extensions.",
+    "extension": "A5 supplied set of coordinate/interval pairs ext(o), compared at fixed coordinate interpretation, resolution and context. Equality is extensional equivalence under that convention; it does not identify historical actors, causal processes or independent random variables.",
+    "jointIndependence": "For declared variables and one probability law, every joint event factors into the product of its marginal probabilities. For a conditional claim, declare the conditioning variable and require factorization in each positive-mass stratum. Pairwise zero covariance, distinct names and disjoint extensions do not imply this.",
+    "namedLineage": "A supplied model label grouping sources by a declared common generation/ancestry mechanism, with membership and any active weighting supplied. Labels and extension distances do not certify that mechanism. Observational identification remains conditional.",
+    "spectralParticipation": "For nonnegative eigenvalues lambda_i of a declared symmetric PSD covariance with positive trace, normalize w_i=lambda_i/trace: K_eff=trace(C)^2/tr(C^2). This spectral measure and a declared lineage-weight measure are distinct objects unless a model connects them."
+  },
+  "id": "extension-lineage-participation@1",
+  "open": [
+    "conditional-lineage-identification",
+    "causal-subgroup-recovery",
+    "oracle-free-reference"
+  ],
+  "reusedEvidence": {
+    "A7/T90": "vectors/laws/a7-t90.json",
+    "A9/T99": "vectors/laws/a9-t99.json",
+    "participation": "dist/measures/error-events.js",
+    "uniformIdentification": "admitted O4 uniform_population_identification, known p and uniform weights only"
+  },
+  "scopeConditions": {
+    "A7": "Fixed coordinates, resolution, context and ceilings; adding live narrowing constraints cannot increase summed finite log-cardinal freedoms. Withdrawal and witnessed local widening can increase freedom on an existing coordinate. A sum over coordinates is not automatically joint Shannon entropy.",
+    "A9": "Representation convention for the object relation; supplies no iteration theorem or independence premise.",
+    "C1": "Historical independence-is-distance wording is not probabilistic factorization; any probabilistic reading needs a separately supplied joint law and independence proof.",
+    "T90": "Nonempty finite admissible states, uniform distribution at fixed resolution: entropy=log2(cardinality). Arbitrary priors, phase interpretations and exclusive lists of freedom-changing acts do not follow. Empty regions need their separately declared algebraic state, not an invented entropy convention.",
+    "T99": "Finite world and records, acyclic reachable dependency graph, total supplied lattice operations. Finite dependency evaluation is not repeated application of an arbitrary monotone map. Reachable ancestral cycles reject; unreachable cycles do not affect the query. General iterative convergence needs a carrier, operator, seed and progress/height hypotheses."
+  }
+}

package/contracts/m0-binomial-bridge.json

@@ -0,0 +1,6 @@
+{
+  "scope": "m0-binomial-bridge",
+  "form": "statement",
+  "status": "prepared-conditional-law-with-kernel-proof",
+  "value": "M0 conditional law: normalized joint product probabilities of named individual Bernoulli outcomes push forward to the exact product-binomial count law. conditional_error_representation and model_mixture_error_representation identify that finite PMF fair-majority expectation and its normalized latent mixture with the existing conditionalError and error for every N. ambient_conditional_error_representation transfers from measurable raw-bit snapshots with that declared joint product law; no binomial count law, expected-error equality or convergence is assumed. fair_loss_average_bridge requires N>0; conditional_error_zero and zero_average_counterexample distinguish N=0. computational_offdiag_limit composes the exact representation with the existing floor_limit_offdiag_mixture theorem under its stated integrability, independence, identical-distribution, mean and measurability assumptions, plus every latent score unequal to 1/2. Domain: finite fixed K, normalized nonnegative fixed weights, p in (1/2,1), c in [0,1), published cumulative-floor allocation and fair ties. Pairwise independence alone is insufficient, certified by pairwise_not_joint. General diagonal, c=1, moment identification, root admission and release readiness are not claimed."
+}

package/contracts/m0-diagonal-conditional.json

@@ -0,0 +1,15 @@
+{
+  "id": "m0-diagonal-conditional",
+  "profile": "lean-kernel@1",
+  "status": "partial",
+  "closed": [
+    "symmetric diagonal subdomain",
+    "actual mean and variance",
+    "member-level Lindeberg",
+    "conditional moving-threshold/fair-tie transfer"
+  ],
+  "generalDiagonal": "open",
+  "unpaidHypothesis": "actual standardized fixed-cut CDF convergence to a continuous symmetric limit",
+  "domain": "published M0 with c<1 and cumulative-floor allocation",
+  "noNewAlgebra": true
+}

package/contracts/m0-keff-interval.json

@@ -0,0 +1,27 @@
+{
+  "casesNotCovered": [
+    "missing/latent activity identification",
+    "causal subgroup recovery",
+    "general oracle-free calibration",
+    "fixed quotas without finite correction"
+  ],
+  "classification": "conditional finite-model coverage and algebra; external calibration still required",
+  "currentCalibration": "No real calibration data or admitted c,p calibration receipts were supplied. Example boxes are synthetic coverage premises only.",
+  "implementationBoundary": "Published cofailurePairs supplies raw observed pair rates. A new public K_eff interval API is not implemented or claimed by this mathematical preparation.",
+  "model": {
+    "attribution": "Observation-position names only under this population-sampling profile: lineage membership is resampled per cell/draw as in the O4 conditional population observation. Fixed source-to-lineage membership is outside this profile without its gamma_N conversion.",
+    "coverage": "P(|Dhat-D|>=epsilon)<=min(1,1/(4T epsilon^2)); joint miss<=min(1,alphaD+betaP+betaC), without requiring calibration/data independence.",
+    "externalCalibration": "Coverage-certified boxes forp,c with allocated marginal error budgets and provenance; neither inferred from agreement nor supplied by these synthetic fixtures.",
+    "frame": "T independent identically distributed complete cells; N>=2 observations per cell, fixed or handed complete frame. No inference of iid from these bytes.",
+    "id": "binary-M0-whole-cell-iid@1",
+    "latent": "Per cell, jointly independent Bernoulli(p) named lineages with declared normalized shares, shared copy probabilityc, fresh member draws. Latent/fresh randomness independent across cells.",
+    "mean": "E U_t=2p(1-p)*(1-c^2*sum(w_l^2)) exactly for finiteN under this observation design.",
+    "members": "Conditional population draws with replacement, independent given the cell latent lineages. Fixed lineage quotas/distinct finite-population member samples require their gamma_N correction and are not this profile.",
+    "optionalCap": "A known deterministic cap is an explicit additional premise. A merely probabilistic cap would require another failure budget and is outside this profile.",
+    "output": "Conditional outer enclosure for K_eff, not exact identified-set equality, latent causal groups, probability of truth or a physical count of independent sources.",
+    "statistic": "U_t=(sum of discordant unordered member pairs)/choose(N,2)=2X_t(N-X_t)/(N(N-1)); Dhat=mean over whole cells. Shared-member pairs need not be independent."
+  },
+  "scope": "m0-keff-interval",
+  "status": "unsigned-preparation",
+  "value": "Under the declared binary M0 conditional-population-iid member design, finite-cell pair discord has the published population disagreement mean. ForT jointly independent complete cells, the all-pairs average is a bounded whole-cell statistic: its tail is at most min(1,1/(4T epsilon^2)), not a bound withT multiplied by the number of pairs. External coverage-calibrated p,c boxes and their union error budgets propagate through the positive M0 denominator to a conditional outer K_eff enclosure. A denominator that may approach zero has no certified finite upper bound unless a valid external cap is declared. Missing calibration is unmeasured; an incompatible confidence box is not an unconditional rejection of the model. No finite interval establishes causal independence."
+}

package/contracts/m0-moment-identification.json

@@ -0,0 +1,42 @@
+{
+  "scope": "m0-moment-identification",
+  "form": "statement",
+  "status": "unsigned-preparation",
+  "classification": "conditional Lean theorem plus exact counterexample",
+  "model": {
+    "id": "binary-M0-uniform-population@1",
+    "latent": "A_l independent Bernoulli(p); joint independence is a declared model hypothesis",
+    "observation": "Conditional on A, independent population draws have correctness probability q=(1-c)*p+c*sum(w_l*A_l); R_j=E[q^j+(1-q)^j]. No truth is needed to count agreement; p is externally declared/calibrated, not inferred from agreement.",
+    "known": [
+      "p in (0,1)",
+      "uniform named shares w_l=1/K",
+      "population expectation semantics"
+    ],
+    "unknown": [
+      "c in (0,1]",
+      "K positive integer"
+    ],
+    "target": "unique(c,K), hence K_eff=K only within this declared uniform family",
+    "units": {
+      "p,c,R_j": "probabilities",
+      "K,K_eff": "dimensionless counts/effective counts"
+    },
+    "samplingBoundary": "Population moments/with-replacement conditional draws. Finite distinct-member observations need their finite population correction; observed moments also need dependent-pair uncertainty. No finite-sample inverse guarantee here."
+  },
+  "value": "Binary M0 with jointly independent Bernoulli(p) latent lineages, known p in(0,1), uniform shares1/K with K>=1 and0<c<=1: the exact population pair and fourth agreement moments identify(c,K). A general quartic expectation bridge and strict monotonicity prove this beyond the old finite grid. For arbitrary normalized weights, p declared is insufficient for this moment map: c=7/10,K_eff=147/4 and c=4/5,K_eff=48 have equal R2,R4 at p=3/4. This collision concerns these moments only, not the entire observable law. R3 remains redundant. No sample uncertainty, causal recovery or contextual activity identification is claimed.",
+  "nullWithoutKnownP": "unmeasured:known-p-required",
+  "nullWithoutUniformity": "identified-set:{normalized positive weights,c satisfying the declared R2/R4 equations}; no singleton inferred",
+  "nullFiniteDataWithoutUncertainty": "unmeasured:population-moment-uncertainty-required",
+  "doesNotClose": [
+    "conditional-lineage-identification",
+    "causal-subgroup-recovery",
+    "oracle-free-reference",
+    "general heterogeneous M0 CLT"
+  ],
+  "priorWorkInPackage": [
+    "Obligations045.M0Coincidence.expect_quadratic",
+    "pair_coincidence",
+    "triple_redundant"
+  ],
+  "newContribution": "Exact quartic model bridge; uniform-family continuous parameter injectivity; explicit nonuniform moment collision."
+}

package/contracts/m1-diagonal-regions.json

@@ -0,0 +1,45 @@
+{
+  "classification": "Lean-checked unbiased diagonal and resolved-state mixture laws; heterogeneous CLT still open; new proof unsigned/unadmitted",
+  "domain": "Same finite named hierarchy, fixed ordered quotas and product primitive law as admitted M1 bridge/offdiag;1/2<p<1,0<=c1,c2<1, freshd>0. No population-resampling or new algebra.",
+  "example": {
+    "ancestors": 1,
+    "c1": "1/5",
+    "c2": "1/6",
+    "diagonalStateMass": "1/16",
+    "fullGeneratorErrorLimit": "1/32",
+    "leaves": 1,
+    "p": "3/4",
+    "provedBy": "Obligations046.M1Diagonal.example_generator_limit_exact",
+    "unbiasedState": [
+      false,
+      false
+    ],
+    "weight": "1"
+  },
+  "id": "m1-diagonal-regions@1",
+  "mainTheorems": [
+    "Obligations046.M1Diagonal.unbiased_conditional_exact",
+    "Obligations046.M1Diagonal.generator_limit_offdiag_or_unbiased",
+    "Obligations046.M1Diagonal.example_generator_limit_exact",
+    "Obligations046.M1Diagonal.diagonal_transfer_of_fixed_cut_limits"
+  ],
+  "model": "binary-M1-ancestor-first-fixed-quota@1",
+  "notClaimed": [
+    "General heterogeneous-diagonal CLT or limit",
+    "Actual primitive CDF convergence derived from variance/Lindeberg alone",
+    "Real causal/model identification or calibration",
+    "New public M1 measurement API",
+    "Native admission/root receipt of this new proof"
+  ],
+  "proof": "proofs/M1Diagonal046.lean",
+  "proved": {
+    "conditionalTransfer": "IF the actual standardized primitive CDF tends at every fixed cut toF continuous at0, then majority error tends toF(0). To claim1/2 also requiresF(0)=1/2. This hypothesis is unpaid for the general heterogeneous diagonal.",
+    "lindeberg": "For fixed epsilon>0, bounded Bernoulli centered tails vanish identically once epsilon^2*variance>=1",
+    "mixtureLimit": "forallA,B: score!=1/2 OR forall i,q_i=1/2 -> full generator error tends to thresholdFunctional",
+    "quotaDrift": "|mu_N-N*S|<=number of named leaves; on diagonal drift^2/variance->0 and moving standardized cut->0",
+    "unbiasedExact": "forall i,q_i=1/2 -> conditional fair-majority error=1/2 for everyN, includingN0",
+    "varianceLower": "N*((1-c1)(1-c2))^2*p*(1-p)<=variance; strict linear growth at fixed parameters"
+  },
+  "sourceSHA256": "sha256:647b435a37369a0128dbbdf6b3c97110293d6bf1c62e63311e346310cfd49463",
+  "value": "In the declared binary M1 ancestor-first fixed-quota model, a latent configuration with every q_i=1/2 has fair-majority error exactly1/2 at every finiteN. Thus the complete implementation and primitive-generator mixture converge to the latent threshold functional whenever every latent configuration is either off diagonal or entirely unbiased. A two-level example p=3/4,c1=1/5,c2=1/6,one ancestor and one leaf has a positive-mass unbiased diagonal state and a Lean-proved full-generator error limit1/32. For heterogeneous diagonal configurations, positive fresh massd=(1-c1)(1-c2) gives variance>=N*d^2*p*(1-p), bounded quota drift, vanishing standardized threshold and the checked bounded-array Lindeberg condition for epsilon>0. These are prerequisites, not a proved general CLT. Transfer from actual primitive fixed-cut CDF limits to the moving majority cut is proved only with that distributional convergence supplied explicitly. The general heterogeneous-diagonal probability limit remains open, as do the three original research cards."
+}

package/contracts/m1-generator-binomial.json

@@ -0,0 +1,26 @@
+{
+  "assumptionBoundary": "Named parent map, jointly independent latent and per-member primitive laws are declared generative hypotheses. Correlated selectors/latent bits, reverse priority and population-resampled memberships are different models.",
+  "classification": "Lean-checked finite probability identity; not admitted",
+  "domain": "Model in M1Bridge046: existing M0 weight/ordered-prefix-quota domain, base.c=c1, named parent map, additional conditional c2 in[0,1); p in(1/2,1), c1 in[0,1). Bridge valid for every finiteN, including algebraic N0; canonical M1 consumer domain stillN>0.",
+  "generator": "FullWorld includes all named latent bits and the three Bernoulli primitive bits for each allocated member. FullLaw uses their normalized nonnegative product masses. No output-binomial premise is assumed.",
+  "id": "m1-generator-binomial@1",
+  "mainTheorems": [
+    "Obligations046.M1Bridge.primitive_branch_fiber",
+    "Obligations046.M1Bridge.primitive_output_is_sequential",
+    "Obligations046.M1Bridge.primitive_generator_to_binomial",
+    "Obligations046.M1Bridge.primitive_law_count_probability",
+    "Obligations046.M1Bridge.fullMass_nonneg",
+    "Obligations046.M1Bridge.fullMass_normalized",
+    "Obligations046.M1Bridge.full_generator_error_bridge"
+  ],
+  "model": "binary-M1-ancestor-first-fixed-quota@1",
+  "proof": "proofs/M1Bridge046.lean",
+  "sourceSHA256": "sha256:a21ac1dedb07af4f28752238edd564293967d8177b75979feead44f01f347b9c",
+  "unpaid": [
+    "M1 off-diagonal limit formalization",
+    "M1 diagonal probability transfer",
+    "Native reader/captures/admission for this proof identity",
+    "Any public M1 API or data-calibrated identification"
+  ],
+  "value": "For the declared binary M1 ancestor-first fixed-quota model, the complete finite generative PMF contains jointly independent named ancestor and leaf correctness bits and three independent primitive bits per member: ancestor selector, conditional leaf selector, and fresh vote. Summing unused primitive draws gives exclusive masses c1, (1-c1)c2, (1-c1)(1-c2)p and (1-c1)(1-c2)(1-p). Pushing this law to named lineage counts yields the product binomial law with q_l=c1 A_parent(l)+(1-c1)c2 B_l+(1-c1)(1-c2)p. For every finite N its fair-majority error is exactly the finite latent mixture of binomial errors. This identity is compiled in Lean without assuming a count law or convergence theorem. It does not prove an asymptotic floor, a CLT, a calibrated public API or causal identification."
+}

package/contracts/m1-offdiag-limit.json

@@ -0,0 +1,42 @@
+{
+  "classification": "Lean-checked conditional probability-limit theorem; unsigned/unadmitted new proof",
+  "derivation": [
+    "q_l in[0,1] derived from the exclusive selector law",
+    "mu_N/N -> S by the existing prefix-quota theorem",
+    "The admitted count bridge plus unchanged product-Bernoulli moment machinery gives Var(total)=sum n_l q_l(1-q_l)<=N/4",
+    "Pointwise loss changes require a deviation beyond the finite mean margin; bound its expected size by Var/(N^2 epsilon^2)",
+    "The positive bound1/(4N epsilon^2) tends0; finite latent mixtures commute with these limits",
+    "The admitted full_generator_error_bridge transfers this implementation limit to the actual full generative PMF"
+  ],
+  "finitePremises": [
+    "Joint product latent and member primitive law are encoded in admitted fullLaw; they are generative hypotheses, not inferred from data.",
+    "N>0 for the finite gap inequality; an eventual condition suffices forN->infinity.",
+    "epsilon>0 fixed; mu_N/N+epsilon<=1/2 for left-side limit1, or1/2<=mu_N/N-epsilon for right-side limit0."
+  ],
+  "id": "m1-offdiag-limit@1",
+  "mainTheorems": [
+    "Obligations046.M1Offdiag.conditional_offdiag_limit",
+    "Obligations046.M1Offdiag.implementation_offdiag_limit",
+    "Obligations046.M1Offdiag.generator_offdiag_limit",
+    "Obligations046.M1Offdiag.example_offdiag",
+    "Obligations046.M1Offdiag.example_generator_limit"
+  ],
+  "model": "binary-M1-ancestor-first-fixed-quota@1",
+  "notAssumed": [
+    "count-law certificate",
+    "probability convergence",
+    "CLT",
+    "an ambient iid array or latent identification oracle"
+  ],
+  "offdiagCondition": "For ALL named ancestor/leaf latent assignmentsA,B: S(A,B)!=1/2. With the declared p domain every latent assignment has positive mass.",
+  "pending": [
+    "Native reader selection/captures/admission for the new off-diagonal proof identity",
+    "M1 diagonal probability transfer",
+    "Original three research demands and empirical model calibration"
+  ],
+  "proof": "proofs/M1Offdiag046.lean",
+  "quantitativeBoundary": "No uniform finiteN cutoff or calibrated empirical confidence interval is claimed. The finite bound requires the displayed mean-margin premise; it is not valid merely because a limiting score is off diagonal.",
+  "scope": "Fixed finite named hierarchy/weights, same p,c1,c2 and prefix allocation as the admitted M1 bridge. Domain1/2<p<1,0<=c1,c2<1. No model or algebra widening.",
+  "sourceSHA256": "sha256:faf6ef0b27049ff4cb89db0e216102239b8bb3f29a76e2a66442fb7c9b6ccd2b",
+  "value": "For the declared finite binary M1 ancestor-first fixed-quota model, if every named ancestor/leaf latent configuration has score S different from1/2, the actual full-generator fair-majority error tends to the exact latent threshold functional sum_z mass(z)1{S(z)<1/2}. The proof uses the admitted finite primitive-generator/binomial bridge, prefix-quota convergence, the exact finite conditional second moment and variance at mostN/4. For any fixed epsilon>0, once the finite conditional mean divided byN lies on the correct side with margin epsilon, its error gap is at most1/(4N epsilon^2). This yields the off-diagonal limit without an assumed probability limit, ambient process oracle or CLT. A concrete shared-parent two-leaf example withp=2/3,c1=c2=1/2 meets the off-diagonal condition. Configurations onS=1/2, causal identification and calibration are outside this theorem; the three original research cards remain open."
+}

package/contracts/m1-subcauses.json

@@ -0,0 +1,66 @@
+{
+  "boundaryReductions": [
+    {
+      "condition": "c1=0",
+      "meaning": "exact finite M0 leaf model with copyc=c2, same weights/allocation/tie; ancestor bits integrate out"
+    },
+    {
+      "condition": "c2=0",
+      "meaning": "ancestor-only model; inherited per-leaf quotas do not necessarily equal prefix allocation of grouped weights at finiteN; no silent exact grouping claim"
+    },
+    {
+      "condition": "c1=c2=0",
+      "meaning": "N independent Bernoulli(p) votes, irrespective of hierarchy"
+    }
+  ],
+  "classification": "declared finite generative model and exact enumerated counterexamples; no admitted law or native receipt",
+  "domain": {
+    "N": "positive integer; fixed source slots allocated to leaves; not O4/O5 population resampling",
+    "ancestors": "Finite distinct named ancestors, fixed total parent map leaf->ancestor; unused ancestors may be omitted",
+    "c1": "0<=c1<1",
+    "c2": "0<=c2<1",
+    "leaves": "Finite ordered distinct named leaves; normalized nonnegative rational weights; positive total1",
+    "p": "1/2<p<1"
+  },
+  "equations": {
+    "a": "c1",
+    "b": "(1-c1)c2",
+    "conditionalCovarianceDistinctMembers": "p(1-p)*(a^2*1{same parent}+b^2*1{same leaf}); distinct members only; fresh terms independent, no cross terms under declared latent joint law.",
+    "d": "(1-c1)(1-c2)",
+    "finiteLaw": "Given latents: X_l~Bin(n_l,q_l), jointly independent; average fairLoss(sumX,N) against exact product binomial law and latentMass.",
+    "latentMass": "product over independent named A_g and B_l of p^bit (1-p)^(1-bit)",
+    "q_l": "a A_parent(l)+b B_l+d p",
+    "score": "S(A,B)=sum_l w_l q_l; d p+a sum_g W_g A_g+b sum_l w_l B_l, W_g=sum_{parent(l)=g}w_l",
+    "thresholdFunctional": "F_M1=sum_z latentMass(z)*(1{S(z)<1/2}+1/2*1{S(z)=1/2}); finite exact functional, not an unconditional limit theorem."
+  },
+  "generativeSteps": [
+    "For each cell independently draw one A_g~Bernoulli(p) per named ancestor and one B_l~Bernoulli(p) per named leaf; all mutually independent conditional onY.",
+    "Allocate n_l(N)=floor(N cumulativeWeight(l+1))-floor(N cumulativeWeight(l)) using the existing ordered prefix-floor rule; sum n_l=N and |n_l-Nw_l|<1.",
+    "For each member independently of all other selector/fresh draws and latent bits: ancestor branch with probabilityc1; else leaf branch with conditional probabilityc2; else fresh Bernoulli(p). Selectors are mutually exclusive branches, not additive causes.",
+    "A member on ancestor branch returns A_parent(l); leaf branch returnsB_l; fresh branch returnsits fresh bit. Decode majority of correctness counts with independent fair tie."
+  ],
+  "hypotheses": {
+    "H-M1-allocation": "Fixed ordered prefix quotas; conditional member independence; whole cells iid only if a repeated-cell uncertainty claim is requested.",
+    "H-M1-diag": "Positive fresh massd and p in(0,1) imply nondegenerate conditional Bernoulli variables; mean driftO(1), variance growsTheta(N). A probability limit1/2 still needs its own CLT/transfer evidence, not an inherited M0 theorem.",
+    "H-M1-joint-law": "All ancestor/leaf latent bits jointly independent Bernoulli(p), conditional onY, with explicit map and order; merely matching marginals insufficient.",
+    "H-M1-offdiag": "For fixed finite named hierarchy and every positive-mass latent state S!=1/2: conditional mean drift bounded byL; variance<=N/4. Proposed |e_N-F| bound sum_z mass min(1,N/(4*(N*abs(S-1/2)-L)^2)) when N*margin>L; otherwise1. Proof/implementation bridge not Lean-checked or admitted for M1.",
+    "H-M1-selectors": "Ancestor-first exclusive branches, c2 conditional; shared selectors/fresh bits excluded.",
+    "endpoints": "d=0 excluded from the general profile. Endpoint finite examples may be studied under a separate explicitly named profile; fair finite tie alone does not guarantee a per-latent limit."
+  },
+  "id": "binary-M1-ancestor-first-fixed-quota@1",
+  "negativeContracts": [
+    "missing-parent-map",
+    "ambiguous-copy-priority",
+    "c2-treated-as-unconditional",
+    "unspecified-latent-joint-law",
+    "population-resampling-as-fixed-quotas",
+    "threshold-functional-presented-as-proved-limit"
+  ],
+  "nullAndImpossible": {
+    "impossible": "No causal subgroup recovery, oracle-free reference calibration, conditional identification, or unqualified floor inferred from coincidence; M0 O5 interval guarantees do not certify M1.",
+    "null": "Declared full hierarchy, normalized weights, p,c1,c2,joint latent law and fixed-quota sampling. Missing any ->unmeasured at contract level; no new public API rejection claimed."
+  },
+  "status": "unsigned-model-preparation",
+  "truth": "Binary Y; fixed per cell or any declared prior. Conditional onY correctness bits below are independent ofY; an actual report equalsY when correct and1-Y otherwise. All probabilities/floors refer to oracle correctness, not observable calibration without truth.",
+  "value": "M1 is a declared binary two-level ancestral-copy model, not an identified causal explanation. In each cell draw mutually independent Bernoulli(p) ancestor and leaf correctness bits. A fixed named parent map assigns each leaf to one ancestor. Each member first copies its ancestor with probability c1; conditional on not doing so it copies its leaf with probability c2, otherwise draws a fresh Bernoulli(p). The exclusive masses are a=c1, b=(1-c1)c2, d=(1-c1)(1-c2), and conditional correctness is q_l=a A_parent(l)+b B_l+d p. Fixed named leaf weights and the existing prefix-floor allocation define independent conditional binomial counts; majority uses a fair tie. The finite latent threshold functional is not by itself a proved asymptotic limit. Off-diagonal convergence has a proposed conditional variance argument; diagonal transfer and model–implementation proof require new M1 evidence. Parent topology, latent joint law, priority and sampling profile belong to the model identity. Missing them is unmeasured, not a discovered cause. The three original research demands remain open."
+}

package/contracts/model-collapse-1a.json

@@ -0,0 +1,16 @@
+{
+  "closure": "Freeze a fully bound run manifest before retrieval/holdout, execute its authorized licensed bytes, report all outcomes and justified uncertainty. Causal interpretations additionally require observable intervention/design.",
+  "id": "model-collapse-1a@1",
+  "negative": "A seed, synthetic table, unnamed dataset, presumed training nonexposure, pairwise iid shortcut or falling finite curve cannot pay the empirical/causal claim.",
+  "nulls": [
+    "dataset identity/license",
+    "trained model and lineage mechanism",
+    "finite hypothesis/effect/error budget",
+    "sampling units and holdout binding",
+    "observable intervention and causal estimand"
+  ],
+  "protocol": "protocols/model-collapse-1a.json",
+  "reason": "No owner-selected licensed dataset or bound training/observation/analysis design supplied. Protocol prepared only; no empirical execution.",
+  "status": "unmeasured",
+  "threeResearchCards": "open; unchanged"
+}

package/contracts/projection-chain.json

@@ -0,0 +1,20 @@
+{
+  "classification": "bounded enumeration; general finite derivation in text, not Lean-certified",
+  "domain": "Finite named X, exact rational mu_X>=0 summing to1; total deterministic maps Y=f(X), Z=g(Y); mu_Y=f_*mu_X and mu_Z=g_*mu_Y, never separately chosen priors. Zero-mass fibers contribute0. Units bits.",
+  "evidenceScope": "All partitions of four named atoms, every nested coarsening, all nonnegative integer masses summing4 divided4. Binary64 log2 tolerance1e-12. This is bounded enumeration, not a universal kernel proof.",
+  "id": "projection-chain-composition@1",
+  "model": "finite-deterministic-chain-same-prior@1",
+  "negative": [
+    "projection/non-chain",
+    "projection/changed-prior",
+    "projection/randomized-map",
+    "mix-loss/prior-required"
+  ],
+  "open": [
+    "General Lean finite entropy proof and native reader admission not prepared by this vector alone",
+    "Original three research cards unchanged/open"
+  ],
+  "policy": "vectors/laws/policy.json",
+  "reuse": "Published mixProjectionLoss with weights0,1 encodes each deterministic projection; same construction as published marginalInformation. No new public API.",
+  "value": "H_mu(X|Z)=H_mu(X|Y)+H_mu(Y|Z). Loss vanishes exactly when the respective projection is injective on positive support. No independence assumption."
+}

package/docs/contracts/experiment1a-public-attribution.md

@@ -0,0 +1 @@
+Publication-attribution correction only. Preserve historical signed sources and receipts. For each path, replace exactly the old bytes by the proposed new bytes: protocols/model-collapse-1a.json sha256:a0fdca5bd239e97258aa5ba9eab3d37afc46d8a1fece767e6c35df0adf7a8ec3 -> sha256:7b8e16cbc8dff5d1654b8bb3710dd619810dde74ad8bb3a70b92dec377af736f; vectors/laws/model-collapse-1a.json sha256:a0778ca34a47aa92fd8c71fa7ca27381cdcfce825988349838dec528563fdca0 -> sha256:dda54cc8ae7d204251fd343f572093434aaee02fd272d5f4f65b5a1ff95e2d50; vectors/laws/experiment1a-protocol.native.json sha256:386ec41df834431492c6e43e644cd9a91a284b967e0bcaf16389ca3cd0709aa8 -> sha256:2b3ddb9e9d08d97d6a33c6726dc8efb298236bb9fd6a0077d0bfdaeda1c16f7b. The1a protocol stays version1 and unmeasured; no dataset retrieved, experimental execution, new model or causal claim. Existing reader/provider/runtime bytes are reused; changed input identity requires a fresh native root reading after authorized admission.

package/docs/contracts/extension-lineage-participation.md

@@ -0,0 +1,13 @@
+# Extension, named lineage and effective participation
+
+Extension is the supplied set of coordinate/interval pairs in A5, compared at fixed coordinate interpretation, resolution and context. Equal extensions identify one extensional class under that convention. They do not identify historical actors, establish a shared cause or prove probabilistic independence. A named lineage is a declared model grouping with a supplied generative/ancestry mechanism, membership and weights. Its name is not observational evidence.
+
+Joint independence is factorization of the declared joint probability law for all joint events. A conditional statement names its conditioning variable and requires factorization in every positive-mass stratum. Different names, disjoint extensions and pairwise zero covariance do not suffice. The fair-bit XOR triple has independent pairs and fails joint factorization.
+
+For finite nonnegative masses with positive total, set w_i=a_i/sum(a) and K_eff=1/sum(w_i^2). It is dimensionless weight participation, between1 and the positive-support size; equal positive weights give that size. It is not a count of causally independent extensions. For a symmetric PSD covariance C with positive trace, normalized nonnegative eigenvalues instead give spectral participation trace(C)^2/tr(C^2). These are different inputs and need a model before they can be equated. Existing participationRatio checks symmetry and nonnegative diagonal but does not certify PSD. A true covariance constructed from the declared error rows supplies PSD; passing arbitrary matrices directly does not. Zero variation is unmeasured. In O4, K_eff=K and identification from population moments apply only in the declared uniform known-p family; nonuniform moment collisions remain.
+
+A7 concerns fixed coordinates, resolution, context and ceilings under added live narrowing constraints. Withdrawal and witnessed local widening may reopen freedom on the same coordinate. A sum of coordinate log-cardinal freedoms is not automatically joint Shannon entropy. T90 equates log-cardinality with entropy for nonempty finite uniformly distributed admissible states at fixed resolution; it does not establish the identity for arbitrary priors or a physical phase interpretation.
+
+A9 is a representation convention, not a convergence theorem. T99 concerns finite records and worlds, an acyclic reachable dependency graph and total supplied lattice operations: evaluation takes finitely many operations. Reachable ancestral cycles reject, while unreachable cycles do not affect a query. An arbitrary monotone iteration still needs its carrier, operator, seed and progress/height hypotheses; a monotone reflection on incomparable singleton intervals has a two-cycle. Different hypotheses and missing proofs are not a fork.
+
+The historical C1 independence-is-distance wording supplies no probabilistic factorization. Definitions settle meaning only. Conditional lineage identification, causal subgroup recovery and oracle-free reference remain open. This companion text changes no signed base, algebra, M/3-M/4 or wire.

package/docs/contracts/installation-046-inputs.md

@@ -0,0 +1 @@
+Native installation closure update only: replace vectors/measures/installed-artifact.cli.json exactly sha256:e13401af6214851b3152214c507cc57788b68cd99b83ad02ffe56684e25bed68 -> sha256:5638dd11e9ac8a3624034048a8fe0fe20ec9e0204d7989d23e0838196643efb8 (2756 bytes). Real Node22/24 clean-consumer probes of admitted-metadata review archive sha256:55963df64637803a4373993bbeb38bd0434237acd74d87a04c4cf2f521329c68 returned PASS closed without ledger/private CAS. Actual test/export/check outputs are retained; the unchanged installed-acceptance/1 reader now returns0..0 in both places and rejects the implementation-negative by name. This does not approve or sign that tarball.1a stays unmeasured and the three research cards stay open.

package/docs/contracts/m0-binomial-bridge.md

@@ -0,0 +1 @@
+M0 conditional law: normalized joint product probabilities of named individual Bernoulli outcomes push forward to the exact product-binomial count law. conditional_error_representation and model_mixture_error_representation identify that finite PMF fair-majority expectation and its normalized latent mixture with the existing conditionalError and error for every N. ambient_conditional_error_representation transfers from measurable raw-bit snapshots with that declared joint product law; no binomial count law, expected-error equality or convergence is assumed. fair_loss_average_bridge requires N>0; conditional_error_zero and zero_average_counterexample distinguish N=0. computational_offdiag_limit composes the exact representation with the existing floor_limit_offdiag_mixture theorem under its stated integrability, independence, identical-distribution, mean and measurability assumptions, plus every latent score unequal to 1/2. Domain: finite fixed K, normalized nonnegative fixed weights, p in (1/2,1), c in [0,1), published cumulative-floor allocation and fair ties. Pairwise independence alone is insufficient, certified by pairwise_not_joint. General diagonal, c=1, moment identification, root admission and release readiness are not claimed.

package/docs/contracts/m0-diagonal-lindeberg.md

@@ -0,0 +1 @@
+For the declared member product, c<1 and the published quota rule imply positive growing actual variance. For each epsilon>0 the member-level Lindeberg truncated-second-moment ratio is eventually exactly zero. At epsilon=0 it equals one for N>0. This is a checked CLT hypothesis, not a CLT theorem.

package/docs/contracts/m0-diagonal-moments.md

@@ -0,0 +1 @@
+Under the declared joint independent Bernoulli member product, the existing sum n_i*q_i and sum n_i*q_i*(1-q_i) equal the actual mean and variance. These identities hold for every finite N. Marginals alone do not justify the variance sum, as the correlated-pair negative demonstrates.

package/docs/contracts/m0-diagonal-symmetric.md

@@ -0,0 +1 @@
+M0 with the published c<1 domain and cumulative-floor quotas. Every conditional q_i=1/2 implies exact conditionalError=1/2 for every N, including zero. A diagonal score alone does not imply every q_i=1/2. General heterogeneous diagonal remains open; c=1 is excluded and parity distinguishes its failure.

package/docs/contracts/m0-diagonal-threshold.md

@@ -0,0 +1 @@
+For the actual M0 product law, moving-threshold majority and fair-tie transfer are conditional on standardized fixed-cut CDF convergence to F, continuous at zero. The tie atom then vanishes; conditionalError tends to F(0), hence to one half only when F(0)=1/2 is declared. The actual M0 CLT remains unpaid and open. A scalar LGN limit alone does not give half error.

package/docs/contracts/m0-keff-floor-refutation.md

@@ -0,0 +1 @@
+Finite refutation under the same declared M0 p=3/4,c=9/10: named weights(1,4,4)/9 and(2,2,5)/9 both have sum(w_l^2)=11/27 and K_eff=27/11,while their static floors are5/32 and1/4. Reuses the unchanged published Obligations045.same_effective_participation_different_floor certificate. Therefore K_eff alone does not determine this floor even with p,c fixed. This does not refute the distinct pair-coincidence identity and supplies no identification or uncertainty guarantee.

package/docs/contracts/m0-keff-interval.md

@@ -0,0 +1 @@
+Under the declared binary M0 conditional-population-iid member design, finite-cell pair discord has the published population disagreement mean. ForT jointly independent complete cells, the all-pairs average is a bounded whole-cell statistic: its tail is at most min(1,1/(4T epsilon^2)), not a bound withT multiplied by the number of pairs. External coverage-calibrated p,c boxes and their union error budgets propagate through the positive M0 denominator to a conditional outer K_eff enclosure. A denominator that may approach zero has no certified finite upper bound unless a valid external cap is declared. Missing calibration is unmeasured; an incompatible confidence box is not an unconditional rejection of the model. No finite interval establishes causal independence.

package/docs/contracts/m0-moment-identification.md

@@ -0,0 +1 @@
+Binary M0 with jointly independent Bernoulli(p) latent lineages, known p in(0,1), uniform shares1/K with K>=1 and0<c<=1: the exact population pair and fourth agreement moments identify(c,K). A general quartic expectation bridge and strict monotonicity prove this beyond the old finite grid. For arbitrary normalized weights, p declared is insufficient for this moment map: c=7/10,K_eff=147/4 and c=4/5,K_eff=48 have equal R2,R4 at p=3/4. This collision concerns these moments only, not the entire observable law. R3 remains redundant. No sample uncertainty, causal recovery or contextual activity identification is claimed.

package/docs/contracts/m0-refuted-inverses.md

@@ -0,0 +1 @@
+Withdrawals preserve history. dark-K-from-error remains withdrawn as a generally single-valued inverse: at p=3/4,c=1/2,uniform K=3 andK=5 have the same static floor1/64,certified by the unchanged published Obligations045.inverse_alias theorem. closed-lineage-law denotes the roadmap WF expected-floor nondecrease proposal and remains withdrawn with the21-state WF witness; it does not denote the valid separately conditioned capacityP0/data-processing statement. No statement here closes conditional-lineage-identification,causal-subgroup-recovery,oracle-free-reference or the actual heterogeneous M0 CLT.

package/docs/contracts/m0-wf-floor-refutation.md

@@ -0,0 +1 @@
+Finite refutation under the published binary M0 static floor functional: p=3/4,c=9/10,neutral Wright-Fisher N=5,initial named counts(3,1,1). All21 multinomial descendant count states,including zero shares,have total probability1. The initial floor32/128=1/4 exceeds E[next floor]=92440/(128*3125)=2311/10000 by189/10000. Reuses the unchanged published Obligations045.wf_transition_normalized and wf_floor_decreases_in_expectation certificates. This refutes universal expected-floor nondecrease; it does not assert universal decrease or a new probability-limit theorem. The floor-monotone-in-t claim is refuted and its closed-lineage-law roadmap alias remains withdrawn; conditional lineage capacity is unaffected.

package/docs/contracts/m1-diagonal-regions.md

@@ -0,0 +1 @@
+In the declared binary M1 ancestor-first fixed-quota model, a latent configuration with every q_i=1/2 has fair-majority error exactly1/2 at every finiteN. Thus the complete implementation and primitive-generator mixture converge to the latent threshold functional whenever every latent configuration is either off diagonal or entirely unbiased. A two-level example p=3/4,c1=1/5,c2=1/6,one ancestor and one leaf has a positive-mass unbiased diagonal state and a Lean-proved full-generator error limit1/32. For heterogeneous diagonal configurations, positive fresh massd=(1-c1)(1-c2) gives variance>=N*d^2*p*(1-p), bounded quota drift, vanishing standardized threshold and the checked bounded-array Lindeberg condition for epsilon>0. These are prerequisites, not a proved general CLT. Transfer from actual primitive fixed-cut CDF limits to the moving majority cut is proved only with that distributional convergence supplied explicitly. The general heterogeneous-diagonal probability limit remains open, as do the three original research cards.

package/docs/contracts/m1-generator-binomial.md

@@ -0,0 +1 @@
+For the declared binary M1 ancestor-first fixed-quota model, the complete finite generative PMF contains jointly independent named ancestor and leaf correctness bits and three independent primitive bits per member: ancestor selector, conditional leaf selector, and fresh vote. Summing unused primitive draws gives exclusive masses c1, (1-c1)c2, (1-c1)(1-c2)p and (1-c1)(1-c2)(1-p). Pushing this law to named lineage counts yields the product binomial law with q_l=c1 A_parent(l)+(1-c1)c2 B_l+(1-c1)(1-c2)p. For every finite N its fair-majority error is exactly the finite latent mixture of binomial errors. This identity is compiled in Lean without assuming a count law or convergence theorem. It does not prove an asymptotic floor, a CLT, a calibrated public API or causal identification.

package/docs/contracts/m1-offdiag-limit.md

@@ -0,0 +1 @@
+For the declared finite binary M1 ancestor-first fixed-quota model, if every named ancestor/leaf latent configuration has score S different from1/2, the actual full-generator fair-majority error tends to the exact latent threshold functional sum_z mass(z)1{S(z)<1/2}. The proof uses the admitted finite primitive-generator/binomial bridge, prefix-quota convergence, the exact finite conditional second moment and variance at mostN/4. For any fixed epsilon>0, once the finite conditional mean divided byN lies on the correct side with margin epsilon, its error gap is at most1/(4N epsilon^2). This yields the off-diagonal limit without an assumed probability limit, ambient process oracle or CLT. A concrete shared-parent two-leaf example withp=2/3,c1=c2=1/2 meets the off-diagonal condition. Configurations onS=1/2, causal identification and calibration are outside this theorem; the three original research cards remain open.

package/docs/contracts/model-collapse-1a.md

@@ -0,0 +1,11 @@
+# Experiment 1a: protocol, no measurement
+
+The versioned model-collapse-1a protocol is prepared independently of dataset selection. No owner-selected dataset, license, bound training/aggregation model or finite analysis design has been supplied; no data were retrieved and no empirical experiment was run. The result is unmeasured, not zero, pass, fail or a law.
+
+Freeze the protocol digest before retrieval. A separate run manifest must bind the owner-selected snapshot/license, model and training provenance, lineage mechanism, volume grid, truth/quality endpoint, sampling units/dependence blocks, exploratory/holdout split, seed, stopping rule, finite hypothesis, effect threshold and valid uncertainty. Freeze its digest before retrieval, and freeze the final exploratory analysis before opening holdout. Digests identify bytes; they do not certify a signed admission or delivery.
+
+Record prior exposure, including this session; unknown training exposure remains unknown. Keep related units and duplicates from leaking across splits. Reuse O5 uncertainty only under its actual sampling/calibration hypotheses. Preserve failed runs and report every prespecified outcome whether the falsifier succeeds or fails. No posthoc endpoint or optional stopping is authorized.
+
+A finite within-lineage curve that continues falling is a candidate empirical challenge only to a specified finite prediction or justified bound. It does not by itself refute a positive asymptotic floor. The training process, quality and lineage assumptions must be supplied; no model is invented from the experiment title. Causal interpretations, including the unspecified historical (ii)/(iii), remain unmeasured without an observable intervention, justified assignment and a defined estimand. Descriptive observations cannot close causal identification.
+
+Conditional lineage identification, causal subgroup recovery and oracle-free reference remain open. This protocol changes no algebra, M/3-M/4, wire, signed base or protocol instrument.

package/docs/contracts/projection-chain.md

@@ -0,0 +1 @@
+Finite named X, exact rational mu_X>=0 summing to1; total deterministic maps Y=f(X), Z=g(Y); mu_Y=f_*mu_X and mu_Z=g_*mu_Y, never separately chosen priors. Zero-mass fibers contribute0. Units bits. H_mu(X|Z)=H_mu(X|Y)+H_mu(Y|Z). Loss vanishes exactly when the respective projection is injective on positive support. No independence assumption. All partitions of four named atoms, every nested coarsening, all nonnegative integer masses summing4 divided4. Binary64 log2 tolerance1e-12. This is bounded enumeration, not a universal kernel proof. General Lean proof and native admission remain pending.

package/package.json

@@ -1 +1 @@
-{"name":"@lapxo/obligations","version":"0.4.5","description":"The algebra of a cell: two poles, origins with phase, eight acts.","type":"module","license":"MIT","author":"Lapxo","repository":{"type":"git","url":"git+https://github.com/Lapxo/obligations.git"},"keywords":["obligations","lattice","provenance"],"sideEffects":false,"engines":{"node":">=22.12"},"exports":{".":{"types":"./dist/lattice/index.d.ts","default":"./dist/lattice/index.js"},"./debit":{"types":"./dist/debit/debit.d.ts","default":"./dist/debit/debit.js"},"./encounter":{"types":"./dist/field/encounter.d.ts","default":"./dist/field/encounter.js"},"./field":{"types":"./dist/field/field.d.ts","default":"./dist/field/field.js"},"./forms":{"types":"./dist/lattice/forms.d.ts","default":"./dist/lattice/forms.js"},"./lattice":{"types":"./dist/lattice/lattice.d.ts","default":"./dist/lattice/lattice.js"},"./measures":{"types":"./dist/measures/index.d.ts","default":"./dist/measures/index.js"},"./packed":{"types":"./dist/debit/packed.d.ts","default":"./dist/debit/packed.js"},"./present":{"types":"./dist/field/present.d.ts","default":"./dist/field/present.js"},"./price":{"types":"./dist/lattice/price/index.d.ts","default":"./dist/lattice/price/index.js"},"./replay":{"types":"./dist/replay.d.ts","default":"./dist/replay.js"},"./shuffle":{"types":"./dist/debit/shuffle.d.ts","default":"./dist/debit/shuffle.js"},"./signatures":{"types":"./dist/field/signatures.d.ts","default":"./dist/field/signatures.js"},"./views/field":{"types":"./dist/views/field.d.ts","default":"./dist/views/field.js"},"./views/product":{"types":"./dist/views/product.d.ts","default":"./dist/views/product.js"},"./package.json":"./package.json"},"files":["dist","README.md","LICENSE","TARGET.bound","receipts.bound","docs/measures.md","samples/measures","vectors/measures","samples/yes/algebra-is-closed.json","samples/yes/field-uses-price-and-debit.json","samples/yes/three-theories-one-algebra.json","src/lattice/index.ts","src/field/motion.ts","samples/yes/installed-artifact.json","samples/no/installed-artifact.json","vectors/measures/installed-artifact.cli.json","docs/img","docs/paper.md","docs/reference.md","docs/learn/in-phase.md","docs/learn/price.md","docs/learn/signature.md","docs/learn/thermometer.md","docs/learn/withdrawal.md","vectors/laws/policy.json","vectors/laws/l7-tolerant-k.json","vectors/laws/l8-phase-defect.json","vectors/laws/l9-mix-loss.json","vectors/laws/implementation-043.json","samples/laws/cofailure-pairs/no/sample.json","samples/laws/cofailure-pairs/yes/sample.json","samples/laws/linage-rest/no/sample.json","samples/laws/linage-rest/yes/sample.json","samples/laws/lineage-ceiling/no/sample.json","samples/laws/lineage-ceiling/yes/sample.json","samples/laws/marginal-information/no/sample.json","samples/laws/marginal-information/yes/sample.json","samples/laws/semilattice/no/sample.json","samples/laws/semilattice/yes/sample.json","vectors/laws/cofailure-pairs.json","vectors/laws/cofailure-pairs.native.json","vectors/laws/coordination.json","vectors/laws/f1/finite-arithmetic.json","vectors/laws/f1/path-witness.json","vectors/laws/helly-comb.json","vectors/laws/helly-comb.lean.json","vectors/laws/linage-rest.native.json","vectors/laws/lineage-ceiling.json","vectors/laws/lineage-ceiling.native.json","vectors/laws/marginal-information.json","vectors/laws/marginal-information.native.json","vectors/laws/meet-semilattice.json","vectors/laws/meet-semilattice.native.json","vectors/laws/resolution-encounters.json","proofs/lineage-ceiling.json","proofs/cofailure-pairs.json","proofs/marginal-information.json","proofs/helly-comb.json","proofs/resolution-encounters.json","proofs/coordination.json","proofs/lean/Obligations/Cell.lean","proofs/lean/Obligations/HellyComb.lean","samples/yes/H1-uncertainty.json","samples/no/H1-uncertainty.json","proofs/m0-045/Obligations045/ConsensusGrid.lean","proofs/m0-045/Obligations045/M0Coincidence.lean","proofs/m0-045/Obligations045/M0Diagonal.lean","proofs/m0-045/Obligations045/M0Limit.lean","proofs/m0-045/Obligations045/M0Offdiag.lean","proofs/m0-045/Obligations045.lean","proofs/m0-045/lake-manifest.json","proofs/m0-045/lakefile.toml","proofs/m0-045/lean-toolchain","vectors/laws/m0-allocation.lean.json","contracts/m0-allocation.json","samples/laws/m0-allocation/yes/sample.json","samples/laws/m0-allocation/no/sample.json","vectors/laws/m0-allocation.native.json","docs/contracts/m0-allocation.md","vectors/laws/m0-finite-grid.lean.json","contracts/m0-finite-grid.json","samples/laws/m0-finite-grid/yes/sample.json","samples/laws/m0-finite-grid/no/sample.json","vectors/laws/m0-finite-grid.native.json","docs/contracts/m0-finite-grid.md","vectors/laws/m0-limit-offdiag.lean.json","contracts/m0-limit-offdiag.json","samples/laws/m0-limit-offdiag/yes/sample.json","samples/laws/m0-limit-offdiag/no/sample.json","vectors/laws/m0-limit-offdiag.native.json","docs/contracts/m0-limit-offdiag.md","vectors/laws/m0-limit-diag.lean.json","contracts/m0-limit-diag.json","samples/laws/m0-limit-diag/yes/sample.json","samples/laws/m0-limit-diag/no/sample.json","docs/proofs/floor-limit-diag.proof.json","vectors/laws/m0-limit-diag.native.json","docs/contracts/m0-limit-diag.md","vectors/laws/m0-pair-coincidence.lean.json","contracts/m0-pair-coincidence.json","samples/laws/m0-pair-coincidence/yes/sample.json","samples/laws/m0-pair-coincidence/no/sample.json","vectors/laws/m0-pair-coincidence.native.json","docs/contracts/m0-pair-coincidence.md","CHANGELOG.md"],"scripts":{"build":"tsc --build","test":"node --import ./test/setup.ts --test --test-concurrency=2 test/*.test.ts test/measures/*.test.ts"},"devDependencies":{"@types/node":"^22","typescript":"^5.9.3"}}
+{"name":"@lapxo/obligations","version":"0.4.6","description":"The algebra of a cell: two poles, origins with phase, eight acts.","type":"module","license":"MIT","author":"Lapxo","repository":{"type":"git","url":"git+https://github.com/Lapxo/obligations.git"},"keywords":["obligations","lattice","provenance"],"sideEffects":false,"engines":{"node":">=22.12"},"exports":{".":{"types":"./dist/lattice/index.d.ts","default":"./dist/lattice/index.js"},"./debit":{"types":"./dist/debit/debit.d.ts","default":"./dist/debit/debit.js"},"./encounter":{"types":"./dist/field/encounter.d.ts","default":"./dist/field/encounter.js"},"./field":{"types":"./dist/field/field.d.ts","default":"./dist/field/field.js"},"./forms":{"types":"./dist/lattice/forms.d.ts","default":"./dist/lattice/forms.js"},"./lattice":{"types":"./dist/lattice/lattice.d.ts","default":"./dist/lattice/lattice.js"},"./measures":{"types":"./dist/measures/index.d.ts","default":"./dist/measures/index.js"},"./packed":{"types":"./dist/debit/packed.d.ts","default":"./dist/debit/packed.js"},"./present":{"types":"./dist/field/present.d.ts","default":"./dist/field/present.js"},"./price":{"types":"./dist/lattice/price/index.d.ts","default":"./dist/lattice/price/index.js"},"./replay":{"types":"./dist/replay.d.ts","default":"./dist/replay.js"},"./shuffle":{"types":"./dist/debit/shuffle.d.ts","default":"./dist/debit/shuffle.js"},"./signatures":{"types":"./dist/field/signatures.d.ts","default":"./dist/field/signatures.js"},"./views/field":{"types":"./dist/views/field.d.ts","default":"./dist/views/field.js"},"./views/product":{"types":"./dist/views/product.d.ts","default":"./dist/views/product.js"},"./package.json":"./package.json"},"files":["dist","README.md","LICENSE","TARGET.bound","receipts.bound","docs/measures.md","samples/measures","vectors/measures","samples/yes/algebra-is-closed.json","samples/yes/field-uses-price-and-debit.json","samples/yes/three-theories-one-algebra.json","src/lattice/index.ts","src/field/motion.ts","samples/yes/installed-artifact.json","samples/no/installed-artifact.json","vectors/measures/installed-artifact.cli.json","docs/img","docs/paper.md","docs/reference.md","docs/learn/in-phase.md","docs/learn/price.md","docs/learn/signature.md","docs/learn/thermometer.md","docs/learn/withdrawal.md","vectors/laws/policy.json","vectors/laws/l7-tolerant-k.json","vectors/laws/l8-phase-defect.json","vectors/laws/l9-mix-loss.json","vectors/laws/implementation-043.json","samples/laws/cofailure-pairs/no/sample.json","samples/laws/cofailure-pairs/yes/sample.json","samples/laws/linage-rest/no/sample.json","samples/laws/linage-rest/yes/sample.json","samples/laws/lineage-ceiling/no/sample.json","samples/laws/lineage-ceiling/yes/sample.json","samples/laws/marginal-information/no/sample.json","samples/laws/marginal-information/yes/sample.json","samples/laws/semilattice/no/sample.json","samples/laws/semilattice/yes/sample.json","vectors/laws/cofailure-pairs.json","vectors/laws/cofailure-pairs.native.json","vectors/laws/coordination.json","vectors/laws/f1/finite-arithmetic.json","vectors/laws/f1/path-witness.json","vectors/laws/helly-comb.json","vectors/laws/helly-comb.lean.json","vectors/laws/linage-rest.native.json","vectors/laws/lineage-ceiling.json","vectors/laws/lineage-ceiling.native.json","vectors/laws/marginal-information.json","vectors/laws/marginal-information.native.json","vectors/laws/meet-semilattice.json","vectors/laws/meet-semilattice.native.json","vectors/laws/resolution-encounters.json","proofs/lineage-ceiling.json","proofs/cofailure-pairs.json","proofs/marginal-information.json","proofs/helly-comb.json","proofs/resolution-encounters.json","proofs/coordination.json","proofs/lean/Obligations/Cell.lean","proofs/lean/Obligations/HellyComb.lean","samples/yes/H1-uncertainty.json","samples/no/H1-uncertainty.json","proofs/m0-045/Obligations045/ConsensusGrid.lean","proofs/m0-045/Obligations045/M0Coincidence.lean","proofs/m0-045/Obligations045/M0Diagonal.lean","proofs/m0-045/Obligations045/M0Limit.lean","proofs/m0-045/Obligations045/M0Offdiag.lean","proofs/m0-045/Obligations045.lean","proofs/m0-045/lake-manifest.json","proofs/m0-045/lakefile.toml","proofs/m0-045/lean-toolchain","vectors/laws/m0-allocation.lean.json","contracts/m0-allocation.json","samples/laws/m0-allocation/yes/sample.json","samples/laws/m0-allocation/no/sample.json","vectors/laws/m0-allocation.native.json","docs/contracts/m0-allocation.md","vectors/laws/m0-finite-grid.lean.json","contracts/m0-finite-grid.json","samples/laws/m0-finite-grid/yes/sample.json","samples/laws/m0-finite-grid/no/sample.json","vectors/laws/m0-finite-grid.native.json","docs/contracts/m0-finite-grid.md","vectors/laws/m0-limit-offdiag.lean.json","contracts/m0-limit-offdiag.json","samples/laws/m0-limit-offdiag/yes/sample.json","samples/laws/m0-limit-offdiag/no/sample.json","vectors/laws/m0-limit-offdiag.native.json","docs/contracts/m0-limit-offdiag.md","vectors/laws/m0-limit-diag.lean.json","contracts/m0-limit-diag.json","samples/laws/m0-limit-diag/yes/sample.json","samples/laws/m0-limit-diag/no/sample.json","docs/proofs/floor-limit-diag.proof.json","vectors/laws/m0-limit-diag.native.json","docs/contracts/m0-limit-diag.md","vectors/laws/m0-pair-coincidence.lean.json","contracts/m0-pair-coincidence.json","samples/laws/m0-pair-coincidence/yes/sample.json","samples/laws/m0-pair-coincidence/no/sample.json","vectors/laws/m0-pair-coincidence.native.json","docs/contracts/m0-pair-coincidence.md","vectors/laws/definitions.native.json","contracts/extension-lineage-participation.json","docs/contracts/extension-lineage-participation.md","proofs/definitions-046/declaration.json","src/laws/laws.bound","vectors/laws/a7-t90.json","vectors/laws/a9-t99.json","samples/laws/definitions/yes/sample.json","samples/laws/definitions/no/sample.json","vectors/laws/experiment1a-protocol.native.json","contracts/model-collapse-1a.json","protocols/model-collapse-1a.json","docs/contracts/model-collapse-1a.md","proofs/experiment1a-046/declaration.json","vectors/laws/model-collapse-1a.json","samples/laws/experiment1a/yes/sample.json","samples/laws/experiment1a/no/sample.json","proofs/m0-046/BridgeNegatives.lean","proofs/m0-046/M0Bridge.lean","proofs/m0-046/Obligations045.lean","proofs/m0-046/Obligations045/ConsensusGrid.lean","proofs/m0-046/Obligations045/M0Coincidence.lean","proofs/m0-046/Obligations045/M0Diagonal.lean","proofs/m0-046/Obligations045/M0Limit.lean","proofs/m0-046/Obligations045/M0Offdiag.lean","proofs/m0-046/lake-manifest.json","proofs/m0-046/lakefile.toml","proofs/m0-046/lean-toolchain","vectors/laws/m0-binomial-bridge.lean.json","contracts/m0-binomial-bridge.json","docs/contracts/m0-binomial-bridge.md","samples/laws/m0-binomial-bridge/yes/sample.json","samples/laws/m0-binomial-bridge/no/sample.json","samples/laws/m0-binomial-bridge/yes/small-binomial-table.json","samples/laws/m0-binomial-bridge/no/pairwise-is-not-joint.json","contracts/m0-diagonal-conditional.json","proofs/m0-046-diagonal/BridgeNegatives.lean","proofs/m0-046-diagonal/M0Bridge.lean","proofs/m0-046-diagonal/M0Diagonal046.lean","proofs/m0-046-diagonal/M0Lindeberg046.lean","proofs/m0-046-diagonal/M0Moments046.lean","proofs/m0-046-diagonal/M0Threshold046.lean","proofs/m0-046-diagonal/Obligations045.lean","proofs/m0-046-diagonal/Obligations045/ConsensusGrid.lean","proofs/m0-046-diagonal/Obligations045/M0Coincidence.lean","proofs/m0-046-diagonal/Obligations045/M0Diagonal.lean","proofs/m0-046-diagonal/Obligations045/M0Limit.lean","proofs/m0-046-diagonal/Obligations045/M0Offdiag.lean","proofs/m0-046-diagonal/lake-manifest.json","proofs/m0-046-diagonal/lakefile.toml","proofs/m0-046-diagonal/lean-toolchain","samples/laws/m0-diagonal-conditional/no/sample.json","samples/laws/m0-diagonal-conditional/yes/sample.json","vectors/laws/m0-diagonal-lindeberg.lean.json","vectors/laws/m0-diagonal-moments.lean.json","vectors/laws/m0-diagonal-symmetric.lean.json","vectors/laws/m0-diagonal-threshold.lean.json","proofs/m0-046-identification/M0Identification046.lean","proofs/m0-046-identification/Obligations045/M0Coincidence.lean","proofs/m0-046-identification/lake-manifest.json","proofs/m0-046-identification/lakefile.toml","proofs/m0-046-identification/lean-toolchain","vectors/laws/m0-identification.lean.json","contracts/m0-moment-identification.json","samples/laws/m0-moment-identification/yes/sample.json","samples/laws/m0-moment-identification/no/sample.json","proofs/m0-046-uncertainty/BridgeNegatives.lean","proofs/m0-046-uncertainty/M0BlockCoverage046.lean","proofs/m0-046-uncertainty/M0Bridge.lean","proofs/m0-046-uncertainty/M0Diagonal046.lean","proofs/m0-046-uncertainty/M0FiniteDiscord046.lean","proofs/m0-046-uncertainty/M0Interval046.lean","proofs/m0-046-uncertainty/M0Lindeberg046.lean","proofs/m0-046-uncertainty/M0Moments046.lean","proofs/m0-046-uncertainty/M0Threshold046.lean","proofs/m0-046-uncertainty/Obligations045/ConsensusGrid.lean","proofs/m0-046-uncertainty/Obligations045/M0Coincidence.lean","proofs/m0-046-uncertainty/Obligations045/M0Diagonal.lean","proofs/m0-046-uncertainty/Obligations045/M0Limit.lean","proofs/m0-046-uncertainty/Obligations045/M0Offdiag.lean","proofs/m0-046-uncertainty/Obligations045.lean","proofs/m0-046-uncertainty/lake-manifest.json","proofs/m0-046-uncertainty/lakefile.toml","proofs/m0-046-uncertainty/lean-toolchain","contracts/m0-keff-interval.json","samples/laws/m0-keff-interval/yes/sample.json","samples/laws/m0-keff-interval/yes/api-rows.json","samples/laws/m0-keff-interval/no/sample.json","vectors/laws/m0-keff-interval.json","vectors/laws/m0-uncertainty-blocks.lean.json","vectors/laws/m0-uncertainty-interval.lean.json","vectors/laws/m0-uncertainty-finite-discord.lean.json","proofs/m1-046-diagonal/M0Bridge.lean","proofs/m1-046-diagonal/M0Diagonal046.lean","proofs/m1-046-diagonal/M0Lindeberg046.lean","proofs/m1-046-diagonal/M0Moments046.lean","proofs/m1-046-diagonal/M0Threshold046.lean","proofs/m1-046-diagonal/M1Bridge046.lean","proofs/m1-046-diagonal/M1Diagonal046.lean","proofs/m1-046-diagonal/M1Offdiag046.lean","proofs/m1-046-diagonal/Obligations045/M0Limit.lean","proofs/m1-046-diagonal/Obligations045/M0Offdiag.lean","proofs/m1-046-diagonal/lake-manifest.json","proofs/m1-046-diagonal/lakefile.toml","proofs/m1-046-diagonal/lean-toolchain","vectors/laws/m1-diagonal-partial.lean.json","contracts/m1-subcauses.json","contracts/m1-diagonal-regions.json","samples/laws/m1-diagonal-regions/yes/sample.json","samples/laws/m1-diagonal-regions/no/sample.json","proofs/m1-046-bridge/M0Bridge.lean","proofs/m1-046-bridge/M1Bridge046.lean","proofs/m1-046-bridge/Obligations045/M0Limit.lean","proofs/m1-046-bridge/Obligations045/M0Offdiag.lean","proofs/m1-046-bridge/lake-manifest.json","proofs/m1-046-bridge/lakefile.toml","proofs/m1-046-bridge/lean-toolchain","vectors/laws/m1-generator-binomial.lean.json","contracts/m1-generator-binomial.json","samples/laws/m1-generator-binomial/yes/sample.json","samples/laws/m1-generator-binomial/no/sample.json","proofs/m1-046-offdiag/M0Bridge.lean","proofs/m1-046-offdiag/M0Diagonal046.lean","proofs/m1-046-offdiag/M0Moments046.lean","proofs/m1-046-offdiag/M1Bridge046.lean","proofs/m1-046-offdiag/M1Offdiag046.lean","proofs/m1-046-offdiag/Obligations045/M0Limit.lean","proofs/m1-046-offdiag/Obligations045/M0Offdiag.lean","proofs/m1-046-offdiag/lake-manifest.json","proofs/m1-046-offdiag/lakefile.toml","proofs/m1-046-offdiag/lean-toolchain","vectors/laws/m1-offdiag-limit.lean.json","contracts/m1-offdiag-limit.json","samples/laws/m1-offdiag-limit/yes/sample.json","samples/laws/m1-offdiag-limit/no/sample.json","vectors/laws/projection-chain.native.json","contracts/projection-chain.json","proofs/projection-chain.md","vectors/laws/projection-chain.json","samples/laws/projection-chain/yes/sample.json","samples/laws/projection-chain/no/sample.json","vectors/laws/definitions.native-capture.json","vectors/laws/experiment1a-protocol.native-capture.json","vectors/laws/m0-identification.native-capture.json","vectors/laws/m0-uncertainty.native-capture.json","vectors/laws/m1-diagonal-partial.native-capture.json","vectors/laws/m1-generator-binomial.native-capture.json","vectors/laws/m1-offdiag-limit.native-capture.json","vectors/laws/projection-chain.native-capture.json","vectors/laws/experiment1a-public-input.native-capture.json","docs/img/logo.svg","docs/img/logo-fork.svg","docs/img/cell.svg","docs/img/withdrawal.svg","docs/img/price.svg","docs/img/thermometer.svg","docs/contracts/m0-moment-identification.md","docs/contracts/projection-chain.md","docs/contracts/m0-wf-floor-refutation.md","docs/contracts/m0-keff-floor-refutation.md","docs/contracts/m0-diagonal-threshold.md","docs/contracts/m1-generator-binomial.md","docs/contracts/m1-offdiag-limit.md","docs/contracts/m0-diagonal-lindeberg.md","docs/contracts/m1-diagonal-regions.md","docs/contracts/m0-diagonal-symmetric.md","docs/contracts/m0-keff-interval.md","docs/contracts/m0-diagonal-moments.md","docs/contracts/m0-refuted-inverses.md","docs/contracts/experiment1a-public-attribution.md","docs/contracts/installation-046-inputs.md"],"scripts":{"build":"tsc --build","test":"node --import ./test/setup.ts --test --test-concurrency=2 test/*.test.ts test/measures/*.test.ts"},"devDependencies":{"@types/node":"^22","typescript":"^5.9.3"}}

package/proofs/definitions-046/declaration.json

@@ -0,0 +1,11 @@
+{
+  "about": "# Extension, named lineage and effective participation\n\nExtension is the supplied set of coordinate/interval pairs in A5, compared at fixed coordinate interpretation, resolution and context. Equal extensions identify one extensional class under that convention. They do not identify historical actors, establish a shared cause or prove probabilistic independence. A named lineage is a declared model grouping with a supplied generative/ancestry mechanism, membership and weights. Its name is not observational evidence.\n\nJoint independence is factorization of the declared joint probability law for all joint events. A conditional statement names its conditioning variable and requires factorization in every positive-mass stratum. Different names, disjoint extensions and pairwise zero covariance do not suffice. The fair-bit XOR triple has independent pairs and fails joint factorization.\n\nFor finite nonnegative masses with positive total, set w_i=a_i/sum(a) and K_eff=1/sum(w_i^2). It is dimensionless weight participation, between1 and the positive-support size; equal positive weights give that size. It is not a count of causally independent extensions. For a symmetric PSD covariance C with positive trace, normalized nonnegative eigenvalues instead give spectral participation trace(C)^2/tr(C^2). These are different inputs and need a model before they can be equated. Existing participationRatio checks symmetry and nonnegative diagonal but does not certify PSD. A true covariance constructed from the declared error rows supplies PSD; passing arbitrary matrices directly does not. Zero variation is unmeasured. In O4, K_eff=K and identification from population moments apply only in the declared uniform known-p family; nonuniform moment collisions remain.\n\nA7 concerns fixed coordinates, resolution, context and ceilings under added live narrowing constraints. Withdrawal and witnessed local widening may reopen freedom on the same coordinate. A sum of coordinate log-cardinal freedoms is not automatically joint Shannon entropy. T90 equates log-cardinality with entropy for nonempty finite uniformly distributed admissible states at fixed resolution; it does not establish the identity for arbitrary priors or a physical phase interpretation.\n\nA9 is a representation convention, not a convergence theorem. T99 concerns finite records and worlds, an acyclic reachable dependency graph and total supplied lattice operations: evaluation takes finitely many operations. Reachable ancestral cycles reject, while unreachable cycles do not affect a query. An arbitrary monotone iteration still needs its carrier, operator, seed and progress/height hypotheses; a monotone reflection on incomparable singleton intervals has a two-cycle. Different hypotheses and missing proofs are not a fork.\n\nThe historical C1 independence-is-distance wording supplies no probabilistic factorization. Definitions settle meaning only. Conditional lineage identification, causal subgroup recovery and oracle-free reference remain open. This companion text changes no signed base, algebra, M/3-M/4 or wire.\n",
+  "at": "witness:obligations046/O8",
+  "by": "target",
+  "form": "alphabet",
+  "measure": "text",
+  "role": "writes",
+  "scope": "prose/en/contracts/extension-lineage-participation",
+  "shape": "docs/contracts/extension-lineage-participation.md",
+  "value": "lock"
+}

package/proofs/experiment1a-046/declaration.json

@@ -0,0 +1,11 @@
+{
+  "about": "# Experiment 1a: protocol, no measurement\n\nThe versioned model-collapse-1a protocol is prepared independently of dataset selection. No owner-selected dataset, license, bound training/aggregation model or finite analysis design has been supplied; no data were retrieved and no empirical experiment was run. The result is unmeasured, not zero, pass, fail or a law.\n\nFreeze the protocol digest before retrieval. A separate run manifest must bind the owner-selected snapshot/license, model and training provenance, lineage mechanism, volume grid, truth/quality endpoint, sampling units/dependence blocks, exploratory/holdout split, seed, stopping rule, finite hypothesis, effect threshold and valid uncertainty. Freeze its digest before retrieval, and freeze the final exploratory analysis before opening holdout. Digests identify bytes; they do not certify a signed admission or delivery.\n\nRecord prior exposure, including this session; unknown training exposure remains unknown. Keep related units and duplicates from leaking across splits. Reuse O5 uncertainty only under its actual sampling/calibration hypotheses. Preserve failed runs and report every prespecified outcome whether the falsifier succeeds or fails. No posthoc endpoint or optional stopping is authorized.\n\nA finite within-lineage curve that continues falling is a candidate empirical challenge only to a specified finite prediction or justified bound. It does not by itself refute a positive asymptotic floor. The training process, quality and lineage assumptions must be supplied; no model is invented from the experiment title. Causal interpretations, including the unspecified historical (ii)/(iii), remain unmeasured without an observable intervention, justified assignment and a defined estimand. Descriptive observations cannot close causal identification.\n\nConditional lineage identification, causal subgroup recovery and oracle-free reference remain open. This protocol changes no algebra, M/3-M/4, wire, signed base or protocol instrument.\n",
+  "at": "witness:obligations046/O9",
+  "by": "target",
+  "form": "alphabet",
+  "measure": "text",
+  "role": "writes",
+  "scope": "prose/en/contracts/model-collapse-1a",
+  "shape": "docs/contracts/model-collapse-1a.md",
+  "value": "lock"
+}