HAC-335: bind the judge package to HAC-343, which is no longer unbound - #32
Merged
Conversation
The package stated, on the root README, the Devpost narrative, the limitations page, the claim ledger and the asset registry, that no evaluation exists: "no SPR, precision, recall, false-block or useful-concurrency number exists in this package, and none is shown." HAC-343 froze a canonical result at 7ede0f9 on 08-19. Every one of those sentences has been false since. Worse, `verify-package.mjs` *enforced* the falsehood — rule 13/14 failed the build if any SPR value appeared next to a number, so the gate would have rejected the correction. The gate is re-pointed rather than relaxed, and the new rule is stricter than the one it replaces. "No number" is trivially checkable. What replaces it needs the frozen export in hand: - every count in a file that renders the comparison must be a frozen `display` value from judge-export.json, so a figure cannot be mistyped, rounded, or recomputed from the raw records; - Panel 1 never appears without Panel 2 in the same file. Panel 1 alone reads as "Interlock is the safe one", and the export forbids exactly that reading — the 0/2 is bounded to COUPLED scenarios and is a property of the evidence being present, not of the engine; - the A3 credibility strip (2/2, 4/4, 2/2) travels with the comparison, so the per-target lock cannot be reduced to a straw man; - every entry in the export's own `mustNotClaim` list is checked against the prose; - the superseded "no SPR exists" sentences fail the build if reverted. Ten negative proofs, one per rule, each asserting its mutation landed. Two contradictions found while binding, both fixed: - the cockpit's A3 strip carried only `2/2 serialized`. That shows the lock ran, not what it cost or what it still missed, which is the half that makes A3 credible. It now binds all three frozen figures. - the reserved HAC-319 shell withheld `SPR` while the panel beside it rendered SPR from HAC-343. A surface that calls a metric withheld next to a panel showing it is not being careful, it is wrong. The shell now withholds precision, recall and fleet-scale behaviour — what is actually still unbound — and names where SPR is bound. HAC-319 proper stays unbound and IL-DIAG-013 stays out of the judge registry with its seam recorded. HAC-343 is its bounded child, not a substitute, and the package now says so instead of letting the bound child imply the unbound parent. Captures re-taken from the changed cockpit; the freshness gate caught them as stale first. pnpm run check exit 0, 644 tests, typecheck and build clean.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
…claration The reserved-shell narrowing changed a source the HAC-334 visual model and IL-DIAG-013 master are generated from, so their committed copies went stale. CI rebuilds and diffs those; the local verifier only checks the committed model against itself, which is why this passed here and failed there. Regenerated, not hand-edited. The master now reads THREE-REGIME EVALUATION NOT YET BOUND and withholds precision, recall and fleet-scale behaviour — no longer SPR, which HAC-343 binds.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The package was stating something false
On the root README, the Devpost narrative, the limitations page, the claim
ledger and the asset registry, this package said:
HAC-343 froze a canonical result at
7ede0f9on 08-19 and is marked Done. Everyone of those sentences has been false since.
Worse, the gate enforced the falsehood. Rule 13/14 of
verify-package.mjsfailed the build if an SPR value appeared next to a number — so the gate would
have rejected the correction.
The gate is re-pointed, not relaxed
The replacement is stricter than what it replaces. "No number" is trivially
checkable; "every number is the frozen one, and none of them travels alone"
needs the export in hand:
displayvalue from
judge-export.json— a figure cannot be mistyped, rounded, orrecomputed from the raw records;
reads as "Interlock is the safe one", and the export forbids exactly that: the
0/2is bounded to COUPLED scenarios and is a property of the evidence beingpresent, not of the engine;
2/2,4/4,2/2) travels with the comparison, sothe per-target lock cannot be reduced to a straw man;
mustNotClaimlist is checked against prose;Ten negative proofs, one per rule, each asserting its mutation landed first.
Two contradictions found while binding
2/2 serialized. That shows the lock ran,not what it cost or what it still missed — the half that makes A3 credible. It
now binds all three frozen figures.
SPRwhile the panel beside it renderedSPR from HAC-343. A surface that calls a metric withheld next to a panel
showing that metric is not being careful, it is wrong. The shell now withholds
precision, recall and fleet-scale behaviour — what is actually still unbound —
and names where SPR is bound.
Boundaries held
HAC-319 proper stays unbound;
IL-DIAG-013stays out of the judge registry withits seam recorded. HAC-343 is its bounded child, not a substitute. No frozen
HAC-343 artifact was modified — the export is read, never written.
Captures re-taken from the changed cockpit; the freshness gate caught them as
stale first.
Verification
pnpm run checkexit 0typecheck,buildclean