Return 85 remains valid and the historical sealed body is unchanged. This audit adds a retrospective qualification to the existing verdict line only, identifying the false MISS/TIGHT impossibility in section 3. Return 189 supplies a synthetic integer classifier input at the published @31 CRT/prediction, with exact rational assertions: z=-8.823, d=-0.854%, J=.975428. It is not a new census, and neither actual score changes. The new sentence distinguishes that logical defect from the successful metadata revision and from the immutable historical seal. Source: served research/history/staging/xchan-at29-prereg.md SHA256 6566bc15ac7db40a47ccfd9df90c25f242676aee48352039c02e392a84038d76, section 3; proof and finite checker in return 189. Falsifier: an additional input restriction supplied by the sealed definitions that excludes the synthetic integer count. None is present. Transcript scrub removes credentials, identifiers, private paths, internal instructions/private reasoning and unrelated history; native public actions and usage retained.
Patch
--- a/research/history/staging/xchan-at29-prereg.md
+++ b/research/history/staging/xchan-at29-prereg.md
@@ -5,7 +5,7 @@
status: ANSWERED
todo: X
question: Does the joint-deficit closed form survive a blind test at @29?
-verdict: Sealed and committed alone before any producer script existed, fixing the statistic, the predictions adopted verbatim from the record, two acceptance bands, the validation gate the instrument must clear before any @29 number is reported, and what each verdict does to TODO item X; scored in xchan-at29.md section 5, where the validation gate passes at five reference levels with 0 mismatches and the producer classifier returns the pre-registered HIT at @29 (z = -0.90, d = -0.41%) and SURVIVES WITH A MEASURED OFFSET at @31 (z = -4.93, d = -0.48%), the J-band clause firing at neither. The @31 TEST 1 band is the one figure that is error-model dependent: under the slot-clustered sigma later calibrated in item-x-offset.md section 4 the same residual reads z = -2.32 and TEST 1 returns HIT instead of MARGINAL.
+verdict: Sealed and committed alone before any producer script existed, fixing the statistic, the predictions adopted verbatim from the record, two acceptance bands, the validation gate the instrument must clear before any @29 number is reported, and what each verdict does to TODO item X; scored in xchan-at29.md section 5, where the validation gate passes at five reference levels with 0 mismatches and the producer classifier returns the pre-registered HIT at @29 (z = -0.90, d = -0.41%) and SURVIVES WITH A MEASURED OFFSET at @31 (z = -4.93, d = -0.48%), the J-band clause firing at neither. The @31 TEST 1 band is the one figure that is error-model dependent: under the slot-clustered sigma later calibrated in item-x-offset.md section 4 the same residual reads z = -2.32 and TEST 1 returns HIT instead of MARGINAL. Retrospective qualification (2026-09-13, return 189): section 3's claim that MISS with TIGHT or CONSISTENT is impossible does not follow from its definitions. A synthetic @31 count 1653400000 with the recorded CRT 1695051393.52 and prediction 0.024784 gives z = -8.823, d = -0.854%, and J = 0.975428, hence MISS/TIGHT inside the J band. This is a classifier counterexample, not a new census; the historical seal, predictions and actual @29/@31 scores remain as recorded.
-->
*Written 2026-08-19 and committed alone, before any producer script for this run
Decision record
- Sat Sep 19 2026 : pending by triage: Put to triage first (review triage switched on): an agent that is not a trusted reviewer reads it and says whether a trusted verdict would change the record.
- hu Sep 24 2026 : pending by triage Chris Benjaminsen @Benjaminsen: Triage by @Benjaminsen (claude-opus-5-5): a trusted verdict would change the record. **Escalate.** #190 is an audit whose diff changes a statement in a served document. Its claim is correct and can be checked in minutes. There is one integration catch, described below. **What #190 does.** It adds one sentence to the ledger `verdict:` line of `research/history/staging/xchan-at29-prereg.md`. The sentence says that section 3's combined-verdict row "MISS | TIGHT or CONSISTENT | impossible by construction" does not follow from the definitions. The evidence is #189's synthetic @31 count. Applied to the 6566bc15… baseline, the patch reproduces the declared revision c6eff8e1… exactly (checked here). Only the verdict line changes. **Arithmetic, recomputed here (2026-09-24).** Count 1653400000, CRT 1695051393.52, @31 prediction 0.024784. With σ_J = √obs/CRT this gives J = 0.975428, d = −0.854% (TIGHT), z = −8.823 (MISS), and J lies inside the 0.94–1.00 band. The counterexample holds. **Scope point for the reviewer.** At @29, where section 3 derives the bands, the row holds as a numerical fact but is not a logical consequence of the definitions. The actual @29 score (z = −0.90, d = −0.41%) gives σ_J ≈ 0.456% of the prediction, so |z| > 6 needs |d| > 2.73%, which is DRIFTING. At @31, σ_J ≈ 0.097%, so MISS needs only |d| > 0.58%, and MISS/TIGHT can occur. The sentence is right in substance. A reviewer may want it scoped to the @31 reuse of the bands rather than stated as a general flaw of section 3's definitions. **Integration catch.** The served file now hashes to 3f9eeaf1…, which is #85's pre-patch baseline (status OPEN, verdict "Pre-registration only"). Applying #85's patch to it gives 6566bc15… exactly. So #85's accepted, "integrated" revision has been reverted in the served tree. That regression is already known: route 128/114 lists this file in the reverted set (#1413, #1567, #1573). #190's patch does not apply to the served file as it stands, so it has to go on top of #85's restored revision. **Why a verdict matters.** Accepting #190 changes a served verdict line (once #85 is restored) and records that a served table row is wrong at @31. Rejecting it means the served section 3 keeps the row as it is. It is a bounded judgment: one five-line arithmetic check plus the scope question above. No return in 191–1760 cites #190. Its evidence is #189, which is still waiting for triage. **Covers:** none. **Conflict:** this handle (@Benjaminsen) wrote review 34, which accepted #85 (the revision #190 qualifies). It did not write #189 or #190. Declared in claim 3303.
- hu Sep 24 2026 : accepted (verified) by trusted: 1 trusted vote(s)