@amityco/social-plus-vise 1.1.0 → 1.1.1

package/CHANGELOG.md

@@ -4,6 +4,13 @@ All notable changes to `@amityco/social-plus-vise` are documented in this file.
 
 The format is loosely based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## 1.1.1 — 2026-06-16
+
+Docs-only patch; no CLI, MCP, rule, or sidecar behavior changes.
+
+### Changed
+- **README Evidence section reworked.** Dropped the pre-1.0 "like-for-like" feed-completeness result (it was measured against an earlier dev build on early-June agent versions); presented the two genuine matched comparisons (SDK compliance, design conformance) as a scannable table; relabeled the former "best case" as **"Feed completeness (capabilities selected)"** and reframed it as *absolute* opt-in completeness rather than a baseline-beating lift; led design conformance with the robust by-name-token signal instead of the fragile one-seed hex anecdote (and dropped the model attribution, given a cross-temporal confound now noted under Boundaries); tightened the limitations and lifted the "stopping condition" framing into the section intro. No benchmark numbers were invented; the SDK-compliance benchmark stays described generically (its internal codename remains denylisted from the shipped surface).
+
 ## 1.1.0 — 2026-06-16
 
 Public GA-readiness pass. No change to `vise check` exit codes or sidecar **formats**. The rule corpus gains three corrected rationales (stale/fictional SDK method names) — a deliberate pre-GA contract reset, with the `test:sidecar-compat` baseline re-frozen accordingly (no published 0.14.x consumers to protect; see below).

package/README.md

@@ -130,13 +130,20 @@ Vise can ingest your aesthetic from an HTML/CSS prototype (`vise design extract`
 
 ## Evidence
 
-Vise's measured effect is on **feed-feature completeness** — does the agent build the whole outcome (pagination, empty/error states, the optional capabilities you asked for) or just the happy path?
+Vise's measured effect is on whether an agent builds the *whole* outcome — pagination, empty and error states, the optional capabilities you asked for — or just the happy path. The durable finding across benchmarks is the **mechanism**, not any single percentage: a checked verification loop beats the same rules pasted into a prompt. What Vise reliably changes is the **stopping condition** — the agent isn't done when the code compiles; it's done when the local contract is green, attested, or blocked on an explicit customer decision.
 
-- **Like-for-like** (pre-registered, K=3 seeds, one firewalled ground truth): the Vise workflow **roughly doubled** feed-feature completeness over the docs-only baseline — to **58–64%** from **21–30%** — and the effect replicated across all three agents (Cursor/Composer 2.5 30→64%, Claude/Sonnet 4.6 27→61%, Codex/GPT-5.4 21→58%), evidence it's a property of the verification loop rather than one model. (Grading was deterministic-only in two of the three cells, which if anything understates the gap.)
-- **Best case** (the Vise arm explicitly opted into the optional capabilities up front; later build): with image/poll/edit selected, the Vise arm reached **97–100%** of the 11-item feed checklist across Cursor/Composer 2.5, Claude/Sonnet 4.6, and Codex/GPT-5.4. This is an opt-in best case, **not** the like-for-like delta above — the baseline arm was not offered the same selection and ran an earlier build.
-- **SDK compliance:** in a greenfield SDK-compliance benchmark (9 slices, single run per cell) the Vise verification-loop arm scored **9/9 (100%)** vs a docs-only baseline at **6/9 (67%)**. The durable finding is the mechanism — a checked loop beats the same rules pasted into the prompt.
-- **Design drift:** under an ambiguous brief (Sonnet, n=3), Vise design runs held hex-literal variance to **0** while the docs-only arm spiked to 15 on one stochastic seed — variance reduction on bad runs, not pixel perfection (a later pre-registered run saw both arms at 0; the robust signal is by-name token usage).
-- **Boundaries:** these are best-case/opt-in comparisons for greenfield social.plus work, self-reported (no third-party audit), not universal claims. Negative results travel with the claims — e.g. no measured advantage on day-2 bug fixing. What Vise reliably changes is the **stopping condition**: the agent isn't done when code compiles; it's done when the local contract is green, attested, or blocked on explicit customer input.
+**Matched comparisons** — Vise's verification loop vs a docs-only baseline, same task and agent:
+
+| Benchmark | What's measured | Docs-only | With Vise |
+|---|---|---|---|
+| **SDK compliance** | correctness slices passed (greenfield; 9 independent slices, single run per cell, two agents) | 6/9 (67%) | **9/9 (100%)** |
+| **Design conformance** | the brand's *named* design tokens used vs hardcoded values (ambiguous brief, n=3) | ~72% | **~90%** |
+
+The mechanism is the durable claim: a checked loop beats the same rules in the prompt.
+
+**Feed completeness (capabilities selected):** when the optional capabilities — image, poll, edit — are selected up front, the Vise arm completed **97–100%** of an 11-item feed checklist across three agents (Cursor/Composer, Claude/Sonnet, Codex/GPT). This is *absolute* completeness from answering Vise's capability questions — **not** a lift over a baseline.
+
+**Boundaries:** these are social.plus's own measurements on greenfield work — self-reported, no third-party audit, measured on earlier builds, and not universal claims. Negative results travel with them: no measured advantage on day-2 bug fixing, and the design-conformance arms were measured at different times (the figure is the robust by-name-token signal, not pixel perfection).
 
 <sub>Cursor, Claude, Codex, GitHub Copilot, VS Code, and other product names are trademarks of their respective owners; social.plus is not affiliated with or endorsed by them. Benchmark figures are from social.plus's own measurements.</sub>
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@amityco/social-plus-vise",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "description": "Skill-guided deterministic CLI for social.plus SDK integration assistance.",
   "license": "SEE LICENSE IN LICENSE",
   "type": "module",