Rank 5 on the docket, and the first item this arc whose expected outcome is
LOSING capability.
conservative_floor is a one-sided Wilson lower bound, and Wilson assumes
independent trials. CORE's pipeline is deterministic, so replaying an identical
case is not a second trial — it is the same trial with a guaranteed outcome.
The deduction sealer folded the raw practice corpus and counted every replay,
so bands recorded 720/720 committed whether that was 720 decisions or 28
decisions seen 26 times each. conservative_floor(720,720)=0.990868 cleared
theta_SERVE=0.99; conservative_floor(28,28)=0.808413 is not close.
RESULT: 25 licensed bands -> 4. Twenty-one demoted.
Survivors: en_conditional_chain, en_disjunctive, en_verb_fact,
en_verb_universal — the only bands whose corpus holds enough independent cases.
What the demotion did and did not do, stated exactly. NO ANSWER CHANGED. No
answer became wrong. `wrong` stayed 0 across all 25 bands. The engine is
exactly as correct as it was. What changed is the CLAIM attached to the answer:
21 bands now serve the same sound conclusion prefixed with "(reasoned, but I
haven't yet earned a verified track record on arguments of this shape)". The
reasoning did not get worse; the boast did.
The producer is hardened, which is the half that stops this recurring.
seal_ledger now refuses outright — assert_sealed_evidence_distinct compares the
ledger about to be written against the corpus's distinct-case count and raises
BEFORE anything is written. The curriculum sealer has carried this guarantee
since ADR-0264 R9; the deduction sealer had no equivalent, which is exactly how
the exposure arose. Catching it at seal time rather than in an audit matters:
an audit finds a padded ledger after it is committed, trusted, and gating a
live flag, and unwinding that then needs a ruling — it needed one.
A HOLLOW GUARD OF MY OWN, caught by sabotage rather than trusted. The first
version compared all_gold_problems() with distinct_gold_problems() — two
functions that agree with each other by construction, and neither of which is
what build_ledger folds. Reverting build_ledger to the raw corpus walked
straight past it AND re-sealed the inflated artifact. Same defect class the
ledger already suffered: a guard checking something adjacent to the thing it
protects. It now takes the BUILT LEDGER as input, because the invariant is
about the artifact. Re-running the identical attack: refused, wrote nothing.
Pinned as test_sealer_refuses_a_ledger_that_counts_replays.
A SECOND MISS, caught by the gate rather than by me. My blast-radius scan
grepped for deduction_serve_license / deduction_serve_ledger / 720 and found
nine files. It missed tests/test_construction_inventory.py, which asserts
deduction SURFACE STRINGS and never mentions the ledger — so a pattern search
over the wrong noun could not see it. The full gate found it. Recorded because
"I scanned for the blast radius" is worth exactly as much as the scan's key,
and mine was the wrong key.
That test is the #138 fabrication defect pin, and it is updated WITHOUT being
weakened. Both surfaces now carry the disclosure prefix, and the fabrication
survives untouched — which is a sharper finding than before: a disclosure
prefix looks like it might cover the hazard, and it does not. The user is still
told they said "furthermore". Exact-match assertions kept on the full strings
so neither the fabrication nor the licence state can drift unnoticed, plus an
isolating assertion that the only difference between clean and leaked recitals
is the fabricated premise.
Four records corrected in the same PR, each of which would otherwise have gone
on asserting capability that no longer exists:
- CAPABILITY_LEDGERS' manifest note ("25 bands at 720/720 wrong=0")
- the published CLAIMS.md claim — caught immediately by G-22's own pin, which
is that fix earning its keep three days after landing
- test_volume_honesty.py's module docstring, which said the exposure was
"NOT silently repaired ... changing it is Shay's ratification, not a
test's". R-13 IS that ratification, so the line went stale by being acted on
- the exposure inventory's framing: it now pins the repair, not the gap
test_volume_honesty's sealed-vs-producer assertion is strictly stronger than it
was. It compared sealed committed counts to the producer's committed counts and
both were the inflated 720 — the assertion held perfectly while the number it
agreed on was the wrong number. Two records agreeing is not evidence that
either is right. It now pins committed == distinct, plus a non-vacuity check.
Nine tests encoded the unearned expectation and were updated to the honest one.
Closes G-19 and H-1.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Wcw2pnMBwyvmNyQg4uPEt4
Two unrelated functions shared the name assert_corpus_sound: the
practice-corpus soundness check in evals/deduction_serve/practice/gold.py
(no args) and the curriculum lane's case-contract check in
evals/curriculum_serve/runner.py (domain, cases). Renamed to
assert_practice_gold_sound and assert_lane_cases_sound respectively so
each name states what it validates, and updated all call sites, the
gold.py __all__ export, and two prose mentions that named the function.
[Verification]: uv run --python 3.12.13 core test --suite smoke -q → 569
passed; --suite deductive -q → 285 passed.
Squashes the arc's work into one commit; the workflow-file edit it originally
carried is excluded (see the end of this message).
## Lane 1 — Workbench recorded a proved answer as ungrounded
With deduction_serving_enabled ratified ON (ADR-0256), workbench/api.py's live
chat route builds a bare ChatRuntime(), so the deduction composer decides
Workbench turns and stamps grounding_source="deduction" — but
_coerce_grounding_source carried a hand-copied whitelist of the six pre-arc
labels and silently rewrote anything else to "none". The runtime comment
reasoned this was inert because "REPL turns do not flow through Workbench's
CognitivePipelineRecord path". True, and irrelevant: the traffic flows the
other way. Stale since 2026-07-24.
Scope is one field. workbench/api.py:818 prefers TurnEvent.epistemic_state,
which read epistemic_state_needed — honest. So the UNregistered path degraded
honestly while the hand-copied whitelist asserted a falsehood; a second copy of
a closed enum was worse than no copy. Hence registration AND derivation:
GROUNDING_SOURCES exposes the Literal's members, and the coercion reads it.
workbench-ui badges/tokens/snapshot follow; enumCoverage.test.ts forces atomicity.
## Lane 2 — the ratification ceremony
The discovery loop was instrumented but not closed. teaching/ratification.py
turns a reviewed decision into a chain record, a corpus commit, and a receipt.
Its design turns on one observation: _ratified_rows DROPS unadmissible rows
silently — correct when serving, a trap when ratifying, because the file grows,
the commit lands, and the band count does not move. So the ceremony refuses to
call an append a ratification until it has re-read the curriculum through the
real loader and seen the chain arrive; a non-admitted append is rolled back.
Validation is a pre-flight courtesy, admission is the proof.
Arena queue entry and ledger reseal are deliberately NOT performed (bridge rule
1); the receipt names them. Front door: `core proposal-queue ratify`, a sibling
of `review` rather than a flag on it.
## Lane 3 — structural closures
- ADR-0263 gains rule 5: absence policy is DECLARED in CAPABILITY_LEDGERS, not
passed at the call site. An AST-matched test fails if a serving path passes
missing_ok again.
- Deductive suite added WHOLE to the pre-push gate: 285 tests in 29s against
smoke's 216 in 62s, so no coverage trade was needed.
- Smoke/CI parity assertion made bidirectional. It was one-directional, and had
drifted.
- test_prior_surface_deduction_binding.py pins correction binding on the
deduction path. The review's diagnosis did NOT reproduce — hash_surface moves
in lockstep — so it pins what is there. Mutation-checked.
- Domain-keyed ADR index over 312 flat-numbered files, explicitly partial.
- Arc-close brief template, plus this arc's own brief filled in against it.
## Lanes 4 and 5 — two premises falsified by measurement, one of them mine
Math 4.2: baseline reproduced (correct=5 wrong=0 refused=495); all four named
cases traced to one seam with each gap isolated by one-variable probes. Then the
number that changes the recommendation: the gap blocking case 0000 affects 1
case in 500, the 'than' gap blocking 0001 affects 2. ADR-0251's prohibition on
per-case growth now rests on a count. No reader change made.
CGA: versor_condition is 0.22% of a turn, not the "~10x proof latency" I claimed
— that multiplied an isolated microbenchmark by a call count and compared it to
a single verdict's latency. The real cost is geometric_product at 33,986
calls/turn (~73%) via cga_inner in search paths. The obvious closed form is NOT
bit-exact (954/4000 in f32); backend.vault_recall's serial fold IS (3000/3000,
worst-rel 0) and is the correct target. cargo test could not run —
static.crates.io is denied by the sandbox network policy — so the Rust parity
question stays open and the typestate lane is carried forward, not shipped
uncompiled.
## Not landed: three lines owed to .github/workflows/smoke.yml
The CI smoke gate is narrower than the local one —
test_pack_draft_serve_boundary.py (ADR-0253 INV-33) has been local-only, unseen
because the parity pin checked one direction. The edit was authored and rejected
at push for lacking the `workflow` OAuth scope, so it is recorded as a named,
dated PENDING_IN_CI exception rather than dropped: the assertion still fires on
any new divergence, and a second guard fires once the three land.
[Verification]: pre-push gates all green — smoke 236 passed, warmed_session 10
passed, deductive 285 passed. Ratification 14, ADR index 5, CLI suites 10.
Grounding/epistemic sweep 741 passed 1 skipped. workbench-ui 598 passed across
73 files, tsc -b clean. capability index 11 passed, digest unchanged. Math
holdout correct=5 wrong=0 refused=495. Committed chain corpora byte-unchanged
after the tests that write to them.
Environment caveat: the repo pins requires-python ==3.12.13, which uv cannot
fetch for linux-x86_64, so `uv sync --locked` fails. All Python runs used a
scratch venv on 3.12.11 with declared deps — not the locked universe, not the
full ~12k suite. The pin was left untouched. Re-run on a 3.12.13 host before
treating this as merge evidence.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FduW6Krm3PPQv3P5iwBYtx
CORE now decides natural-English deductive arguments end-to-end:
'If it rains then the ground is wet. It rains. Therefore the ground
is wet.' -> ENTAILED, rendered over the user's own clauses. Both of
ADR-0256's documented boundary cases (multiword conditional, nested-
negation contraposition) are now decided; their lane gold updated
declined->entailed.
- generate/proof_chain/english.py (new): closed function-word grammar
over OPAQUE clause-atoms (minted ids); if/then, [either] or, and,
three negation forms (leading not / sentential / copular incl.
contractions); typed refusals; honesty caps. Soundness: positive
verdicts are substitution-closed; UNKNOWN is scoped + guarded
(quantifier-led, is-a membership, unnormalizable negation refuse
out of the opaque band).
- render_entailment_english: verdicts quoted back verbatim; UNKNOWN
phrasing explicitly scoped to the opaque reading.
- chat/deduction_surface.py: v2-EN fallback tier AFTER v1/v1b —
monotone widening, byte-identical on previously-served arguments.
- Four en_* shape-bands EARN SERVE via the ADR-0199 arena (720/band,
wrong=0, reliability 0.99087 >= 0.99); 9-band ledger re-sealed.
- evals/deduction_serve/v2_en: 26 hand-authored real-English cases
(content disjoint from the synthetic lexicon), wrong=0; report
re-pinned.
- realizer guard (ADR-0075): deduction surfaces exempt — quoted
templates are not slot-composed articulations (pack 'open'=VERB
was rejecting an honest quoted 'the door is not open'); pinned
cold+warm in e2e.
[Verification]: deductive 84 passed; smoke 180 passed; cognition 122
passed 1 skipped; arena 9x720 wrong=0 all SERVE; lanes v1 28/28,
v2_en 26/26.
Deduction serving stops being a bare flag and becomes an EARNED capability
governed by the ADR-0175 reliability gate over the ADR-0199 learning arena
-- the arena's second concrete instance (GSM8K math is the first) and the
reliability substrate's first non-estimation serving consumer, a concrete
dent in that zone's standing 'designed, not wired' critique.
Non-circular thesis: the ROBDD engine is sound+complete (never wrong on the
problem it's handed), so the license does NOT certify it. It certifies the
full pipeline (reader -> projector -> engine) per argument shape -- the
FALLIBLE part is the template reader, which can misparse an argument and
hand the engine the wrong problem. The arena catches that as a 'wrong' by
comparing the pipeline's committed outcome to by-construction gold.
New: generate/proof_chain/shape.py (4 exhaustive structural shape-bands, the
capability axis, shared by arena + serving), evals/deduction_serve/practice/
(the deduction arena instance: deterministic synthetic corpus, by-construction
gold independent of the reader per ADR-0199 L-2, cross-checked against the
INDEPENDENT truth-table oracle before sealing), chat/data/deduction_serve_
ledger.json (committed SHA-sealed ledger, 4 bands x 720 correct/0 wrong),
chat/deduction_serve_license.py (tamper-evident serving reader, mirrors
estimation_license.py), tests/test_deduction_serve_license.py (13 tests),
docs/adr/ADR-0256 (+ resolves the ADR-0206 numbering collision in doc form).
Changed: chat/deduction_surface.py (composer consults the license: earned
band -> authoritative Phase-1 surface; unearned/stripped ledger -> same sound
answer served DISCLOSED/hedged -- authority now rests on committed evidence,
not a boolean), core/config.py (flag docstring: enables an EARNED path),
core/cli_test.py (deductive suite runs the license test).
All four structural bands earn SERVE at reliability 0.99087 (>= theta_SERVE
0.99) with wrong=0, so Phase-1 behavior is preserved byte-for-byte; the
gate's teeth are proven by injecting an empty ledger (the same sound answer
degrades to a disclosed hedge). Did NOT reuse the ADR-0206 govern_response
bridge (its STRICT/APPROXIMATE 'widen-past-strict' semantics are the opposite
shape from deduction's always-sound answer); did NOT rewrite report.json's
stale adr field (SHA-pinned bytes, documented in ADR-0256 s2a instead).
[Verification]: smoke 180 passed; cognition 122 passed/1 skipped;
core test --suite deductive 38 passed; architectural_invariants 75 passed;
practice runner 4 bands all SERVE wrong=0.
Claude