# Adversarial review log — partial-reprogramming-human-aging

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).

| Item | Value |
|---|---|
| Level before → after | Γ (Mechanistically plausible, partial system support) → Γ (unchanged) |
| Confidence before → after | Moderate → Moderate (unchanged) |
| Role / Momentum / Stability / Divergence | unchanged (reinforcing / stabilizing factor; Decisive work underway; Uncertain; Contested) |
| Findings | Critical 0, Major 1, Minor 0, Editorial 0 |
| PMIDs verified | 69/69 reference PMIDs resolved to the stated title, first author, journal and year; no retractions or expressions of concern in the retrieved records |
| Load-bearing claims spot-checked | 13 |
| Validator | PASS — 0 fail, 0 warn, 30 pass (validate_review.py v2.3); output in `validator.txt`, pre-review report kept at `sol61-original/validation-report.json` |

## Findings

| ID | Severity | Location | Original → corrected | Evidence |
|---|---|---|---|---|
| PR-1 | Major | AT A GLANCE, "Best-supported rival" | "Reprogramming reverses causal aging processes" → "Molecular reset and function are partly separable (tournament rank 1; the nominal model ranks 2)" | Score card `competing_models.models_evaluated`: that model is `M1`, `is_nominal: true`, `rank_in_tournament: 2`, level Γ. The preregistration lists it first as the nominal. The rank-1 non-nominal model is `M2` ("Molecular reset and function are partly separable", Ε), which is the model the paragraph's own evidence supports. The renderer took `models_evaluated[0]` — the nominal — as the rival. |

## Rejected findings (checked, no change)

1. Reference years 3, 47, 66, 69 differ from PubMed by ±1 (print/issue vs e-pub) — within convention.
2. PubMed fields dropping italicised gene names (Vitamin B12, p16High, OSK symbols) — PubMed artifact, not a misstatement.
3. Momentum "Decisive work underway" — artifact present and verified against the local registry record for NCT07290244 (recruiting, phase 1, non-randomized, unmasked, 18 estimated, primary completion 2027-05, `hasResults: false`).
4. Divergence "Contested" with null implied level — permitted where no single implied level applies; basis stated.
5. Aggregation — primary-subclaim on SC2 gives Γ; weakest-link and scope-matched also Γ; SC5 (Ε) is scope-restricted to preclinical recipients and cannot set the human-scoped headline.
6. Headline level Γ — re-derived. Δ would require a field-valid causal perturbation for the claimed (adult-human clinical healthspan) scope; the only human evidence is an uncontrolled open-label 3-recipient sponsor interim report. Β would be wrong because controlled animal perturbations with functional readouts exist. Γ stands.
7. Numeric claims verified: 124-week start and ~8.9 vs 18.5 weeks remaining median life (PMID 38381405, +109%); ~30-year transcriptome rejuvenation (PMID 35390271); TET1/TET2 requirement (PMID 33268865); full reprogramming without c-Myc plus teratomas and integration (PMID 29985406); unchanged observed tumour incidence (PMID 39259812); 20 MCI participants, stable cognition, uncorrected CBF (PMID 37994989); up to 9 years replicate noise (PMID 36277076).

## Residual limitations

- COMPETING MECHANISTIC MODELS is ordered by model ID, not tournament rank, so the rank-1 model prints second. Left in place (renderer behaviour shared with the sibling cases of this batch).
- Access depth unchanged: 5 partial-body and 64 abstract/status examinations; claims resting on Methods or Supplementary detail are verified only to that depth.

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