# Adversarial review log - pku-phenylalanine (9 October 2026 revised edition)

Authoring model: **Sol 6.1**. Adversarial reviewer: **Claude Opus 5.5 (Anthropic)**. Date: **2026-10-14**.
Round 2 of the Sol 6.1 HSEF adversarial pass. This is a revised edition, so findings are marked as
revision-introduced (regression) or inherited.

| Item | Value |
|---|---|
| HSEF level before -> after | Theta (High predictive accuracy, minimal unresolved exceptions) -> Theta (unchanged) |
| Confidence before -> after | High -> High (unchanged) |
| Causal role | primary driver (unchanged) |
| Findings | Critical 0 - Major 0 - Minor 3 (2 of them revision-introduced) - Editorial 0 |
| PMIDs verified | 27/27; 0 failures; no retraction or concern flag on any record. The one cross-referenced non-ledger PMID (29653686) also checked and correct. |
| Claims spot-checked | 19 load-bearing numeric claims against original abstracts or full text; every figure matched |
| Validator | **PASS** - 0 fail, 2 warn, 35 pass. Both warnings are the disclosed `v5.calibration` unblinded-calibration warnings, which the review itself reports as "0 fail / 2 disclosed warn". |

## Findings

| ID | Severity | Location | Correction |
|---|---|---|---|
| AR-1 | Minor - revision-introduced regression | Spliced Consensus block, closing paragraph; same text in `consensus-addendum.md` | Rescoped the false-approval clause. It asserted "the earlier approval" where no earlier adversarial approval exists for this case. Rescoped, not deleted. |
| AR-2 | Minor - revision-introduced regression | Spliced Consensus block, "Frozen inventory" paragraph; same text in `consensus-addendum.md` | "retained as raw JSON in this folder" -> "under `checks/pku-phenylalanine/` in the audit root", where the four files actually are. |
| AR-3 | Minor - inherited | Citation ledger preamble | Stale pass-01 sentence "All 15 numbered sources are primary research reports" replaced with the correct 27-source, 25-Tier-A statement, matching the header, the table and the pass-02 note in the same section. |

No level, confidence, role, momentum, stability, divergence, subclaim level, source count or claim count changed.

## Rejected findings

Theta re-argued from both sides and retained (the untreated-phenotype exception is to penetrance of the disability
threshold, not to the causal effect, and is already the named Omega blocker; the residual-deficit blocker is
answered by source [3]'s own best-treated subgroup and the authors' own causal attribution, which the review
correctly labels observational); Omega correctly withheld; the four-modality independence basis survives the v5.2
shared-failure test, and the pharmacological-lowering row is correctly excluded as non-qualifying; Momentum
"Mechanistic only" was actively tested against the blind run's "Decisive work underway" and the registry check
makes the rejection sound; Confidence High untouched by the lead's two-step test because its rationale never
leaned on access depth; the stacked pass-01/pass-02 presentation with "superseded where it concerns" pointers is
deliberate and internally consistent; the "prior H. pylori deliverables" sentence in Reviewer notes appears in no
other review in the audit and so is this session's own housekeeping, not mis-spliced boilerplate; no
collective-author and no missing-space rendering defects exist in this document; Rubric 06 was declined with a
stated reason rather than applied literally.

## Note for the lead

This is the strongest-written review in this group. The Oct 9 splice did **not** produce the known REVIEWER
NOTES contradiction here, because this review's notes never claim that no Consensus work was performed. It did
produce two provenance defects, both now rescoped (AR-1, AR-2). The false-approval clause in this case uses a
different wording from the variant the lead circulated - "the earlier approval does not certify this supplement"
rather than "The earlier adversarial approval refers to its original edition". A corpus sweep found that variant
in 31 revised reviews and the circulated variant in 168, so both templates are in play and the lead's sweep
should match on either.

## Files changed

- `pku-phenylalanine.review.md` (banner, ADVERSARIAL REVIEW AND APPROVAL section, AR-1/2/3)
- `consensus-addendum.md` (AR-1, AR-2)
- `validator.txt` (new)
- `sol61-original/` holds the pre-review copies.

**Approval:** Following adversarial review and the corrections listed above, the revised document is approved. - Claude Opus 5.5 (Anthropic), 2026-10-14.