# Adversarial review log — excitotoxicity-als (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 adversarial review programme (group 3). Canonical copy reviewed:
`consensus_consistency_audit_2026-10-09/revised/excitotoxicity-als/`.

| Item | Value |
|---|---|
| Level before → after | Γ (Mechanistically plausible, partial system support) → Γ (unchanged) |
| Confidence before → after | Moderate → Moderate (unchanged) |
| Momentum before → after | Decisive work underway → **Mechanistic only** |
| Role / Stability / Divergence | unchanged (reinforcing / stabilizing factor; Stable; Contested) |
| Findings | Critical 0, Major 1, Minor 3, Editorial 0 |
| PMIDs verified | 69/69 reference PMIDs resolved to the stated title, first author, journal and year; no retractions or expressions of concern; the two linked errata are correctly attached to parent sources rather than numbered |
| Load-bearing claims spot-checked | 12 |
| Validator | PASS — 0 fail, 1 warn, 30 pass, using the baseline validator recorded in `baseline-validation.json` (sha256 e91528f6…). The single WARN (`v5.claim_form` "phenomenon") is identical to the pre-review baseline and is a standing rejected finding. Output in `validator.txt`; pre-review report kept at `sol61-original/validation-report.json` |

## Findings

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| EX-1 | Major | AT A GLANCE table, "Momentum:" line, score card `extended_scoring_MSD.momentum` | "Decisive work underway" → "Mechanistic only" with an artifact naming specific work | Momentum requires an artifact naming the work, programme or trial type, and "Decisive work underway" means someone is running an experiment that could cross the next threshold. The original artifact named no study and disclaimed any recruiting trial, while the review's own decisive-experiment section says the target-engaged sporadic-human dependency test does not exist. Momentum modifies no letter. |
| EX-2 | Minor (regression) | Consensus provenance reconciliation block; `consensus-addendum.md` | "The earlier adversarial approval refers to its original edition" → "No earlier adversarial approval is attached to this case" | Neither edition carries an adversarial section or log; the sentence is unconditional boilerplate from `integrate_consensus.py`. The earlier edition does not contain the claim, so the revision introduced it. |
| EX-3 | Minor | Falsification and pitfalls | "and32484501" → "and 32484501" | Missing space before a numeral (known rendering defect). |
| EX-4 | Minor | REFERENCES entry 11 | "DOI: ." → "No DOI in the PubMed record." | Empty DOI field (known rendering defect); PubMed carries no DOI for PMID 11357955. |

## Rejected findings (checked, no change)

1. SC2 at Ζ — `independence_basis` names four lines and the distinct systematic error each excludes, as v5.2 requires; the subclaim scope is forced-lesion sufficiency, and the record itself discloses that the glial line does not exclude developmental or astrocytic confounds. Debatable, not forced by the framework; SC2 does not gate the headline.
2. Aggregation — scope-matched correctly returns SC4 (Γ), the highest subclaim whose scope matches the sporadic-human-primacy claim; the narrower Ζ/Ε subclaims are reported separately, and both alternative rules are reported (weakest-link Β, primary-subclaim Γ).
3. Validator WARN `v5.claim_form` "phenomenon" — the registered claim states the mechanism plus its consequence; standing rejected convention.
4. Reference 16 (2022 vs PubMed 2021) — ±1 print/e-pub, already flagged in the timeline.
5. PubMed title fields dropping italicised gene names (STMN2 in reference 49) — PubMed artifact.
6. Batch-wide revision artifact reported by the lead (spliced Consensus block contradicting preserved REVIEWER NOTES) — checked and **not present** in this case: the preserved methods and reviewer notes affirm signed-in browser Consensus use and state only that zero Consensus *API* calls were made. No scoping edit needed. `review_metadata.reviewer` ("Codex") left unedited as house provenance.
7. Headline Γ re-derived: Δ would need a disease-relevant causal perturbation at the sporadic-human scope (absent; the selective human tests are null or harmful), and Β would ignore the endogenous lesion/rescue work. Γ stands.
8. Numeric claims verified: PMIDs 25297012 (513 randomized, 340/173, null stage 3), 34191081 (66/65, 7/21 vs 18/22, −8.4 with differing abstract and Table 2 intervals), 8302340 (155), 8676624 (959, p=0.076 unadjusted vs 0.002 adjusted), 19961264 (59, non-significant), 33226425 (65), 28342750 (progression and lifespan extended, onset unchanged), 29458024 (Hoxb8-Cre GLT1 deletion, abstract only).

## Residual limitations

- The spinal GLT1 lesion carrying part of the SC2 independence basis is abstract-only; its controls are unaudited.
- Linked errata PMID 38890465 and PMID 36480942 are recorded but unexamined.
- Access depth unchanged: 6 partial-body and 63 abstract/status examinations.
- The divergent earlier edition at `hsef_v53_popularity_ranks_31_35_2026-10-08/excitotoxicity-als/excitotoxicity-als.review.md` was **not** edited, per instruction. It still carries findings EX-1, EX-3 and EX-4 (but not EX-2, which the revision introduced).

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