{
  "case": "resident-brain-microbiome",
  "edition": "2026-10-09 revised edition (consensus_consistency_audit_2026-10-09/revised)",
  "authoring_model": "Sol 6.1",
  "adversarial_reviewer": "Claude Opus 5.5 (Anthropic)",
  "date": "2026-10-14",
  "level_before": "Α",
  "level_after": "Α",
  "confidence_before": "Moderate",
  "confidence_after": "Moderate",
  "momentum_before": "Decisive work underway",
  "momentum_after": "Mechanistic only",
  "pmids_checked": 59,
  "pmids_failed": 0,
  "claims_spot_checked": 12,
  "full_texts_reread": [
    "40284643",
    "33771222",
    "23355888",
    "28676754",
    "29881346",
    "39292785"
  ],
  "validator": "framework/5.4-registry-compatibility: FAIL - 1 fail, 1 warn, 33 pass; byte-identical result on the pre-edit copy, so the v5.lay_summary FAIL (a v5.4-only requirement a v5.3 review cannot satisfy) and the v5.claim_form phenomenon WARN are pre-existing and unaffected",
  "crux_verification": {
    "question": "Does the review's treatment of negative/blank controls and of the contamination counter-explanation match what the cited papers report?",
    "answer": "Yes on every point checked, and in each case the review takes the conservative reading.",
    "details": [
      "40284643: '86 samples from 31 specimens, 14 AD/17 controls, four extraction blanks' matches the Methods verbatim ('Four blank extraction controls ... without sample input'; decontam + microDecon). The paper's own abstract says 30 specimens while its Methods say 31 (14+17); the review follows the Methods and is correct. The review also correctly refuses the paper's title claim of 'Absence of Postmortem Contamination': no significant edge-core difference is not an equivalence test and does not exclude reagent, vascular or intratissue preanalytic sources.",
      "33771222: '47 postmortem and three surgical brains' matches (N=47 postmortem, PD 25/HC 22; n=3 fresh cortex from living volunteers at resective surgery). The positive-control limitation is accurate: the mock recovered only six of eight expected species, Listeria monocytogenes absent and Bacillus subtilis under-recovered, no mechanical lysis. Used to bound the null in both directions, not to dismiss it.",
      "23355888: heated/unheated assignment correct - the original states ODC13 and ODC14 homogenates were implanted with and without prior heat treatment respectively - so the donor-confound objection is a real design limitation. Seven-week RNA recovery without neurological deficit as reported.",
      "28676754: the paper itself states it was 'carried out with normalized, re-amplified libraries and therefore was not designed to assess actual bacterial numbers', and discusses human 18S co-amplification. Both limits reproduced.",
      "29881346: 'multiple PCR negatives and sterile rapid-autopsy handling' supported - separate rooms/glassware, aliquoted reagents, positive-displacement pipettes, aerosol-resistant tips, multiple negative controls; sterile instruments in a laminar flow cabinet; rapid neuropathological autopsy, mean postmortem interval 4.5 h. The review's hedge that SEQUENCED bacterial extraction blanks remain unverified is accurate: the retained passages document PCR negative controls, not sequenced blank libraries.",
      "39292785: perfusion with haemoglobin verification, lysis-buffer-only negative plates, CSF from the same animals plated identically, environmental taxa not matching the fish isolates, 54 brain isolates, tagged Plesiomonas significantly lower at day 14 than at 3 hours, comparative genomics on isolates 'from two different laboratory rainbow trout' - all as stated."
    ]
  },
  "findings": [
    {
      "id": "RBM-1",
      "severity": "Major",
      "regression": true,
      "location": "Consensus provenance reconciliation - 9 October 2026, closing sentence; same sentence in consensus-addendum.md",
      "original": "The earlier adversarial approval refers to its original edition",
      "corrected": "No earlier adversarial approval is attached to this case",
      "evidence": "The earlier edition carries no ADVERSARIAL REVIEW AND APPROVAL section and no banner, and no ADVERSARIAL_REVIEW.md or adversarial-findings.json in the deliverables names this case. Spliced in unconditionally by the Oct 9 integration step and absent from the earlier edition, so a revision-introduced regression."
    },
    {
      "id": "RBM-2",
      "severity": "Major",
      "location": "AT A GLANCE table; Final score / M-S-D section; scorecard extended_scoring_MSD.momentum",
      "original": "Momentum: Decisive work underway (artifact: 2023 full-length profiling, 2024 salmon culture, 2025 edge/core; 'no current recruiting trial asserted')",
      "corrected": "Momentum: Mechanistic only, with the artifact rewritten to state why",
      "evidence": "Framework: 'Decisive work underway - Someone is running an experiment that could cross the next threshold', with a required artifact naming the work, programme or trial type. The artifact named only completed publications and then conceded that no recruiting trial is asserted. Under lead policy a completed publication or trial never satisfies this value. Momentum modifies no letter, so Α/Moderate are unchanged."
    },
    {
      "id": "RBM-3",
      "severity": "Minor",
      "location": "scorecard final_score.confidence_rationale (3 occurrences)",
      "original": "density, heterogeneous controls and incomplete access set the confidence, not correctness",
      "corrected": "re-derived from evidence quality alone; access depth struck; band stays Moderate",
      "evidence": "Lead policy two-step test. Remaining determinants are substantive and documented: experimentally demonstrated artifact mechanisms, no human study excluding reagent/host/vascular/preanalytic origins together, a host-aware null bounded by its own positive-control recovery failures, non-independent multiple samples per donor, and the strongest residency positive sitting in one teleost platform. Moderate stands; not raised to High."
    },
    {
      "id": "RBM-4",
      "severity": "Editorial",
      "location": "Competing models inline citation runs (6 prose sites, 8 scorecard sites)",
      "original": "PMID33771222,25387460,31164452,31239396",
      "corrected": "PMIDs 33771222, 25387460, 31164452, 31239396",
      "evidence": "Known rendering defect (missing spaces around numerals). No citation added, removed or re-ordered."
    }
  ],
  "rejected_findings": [
    "'Best-supported rival: Low-biomass reagent/host artifact' - M2 has is_nominal=false and rank 1; the nominal M1 ranks fifth. The nominal-as-rival renderer defect is NOT present.",
    "Primary-subclaim aggregation returning Α - SC1 is the registered existence gate and is scope-matched; weakest-link and scope-matched sensitivities also return Α.",
    "REVIEWER-NOTES contradiction - NOT present. Both the methods and reviewer-notes sections record browser Consensus use (two Pro queries, zero API/Deep) consistent with the spliced provenance section.",
    "The 30-vs-31-specimen discrepancy in PMID 40284643 - the review's 31 follows the paper's Methods and is correct; the paper's own abstract is the inconsistent text.",
    "Confidence Moderate paired with an Α level - the framework states the two axes are independent; 'we are confident the existence gate is not yet crossed' is the intended reading.",
    "Role of mechanism 'epiphenomenon' - derivation step 2 correctly applied; not taken from tournament rank.",
    "Stability 'Stable' - no named pending trigger asserted, which the framework says is stability by definition.",
    "Collective-author rendering defect - none in this reference list.",
    "Headline counts - 59 sources / 63 claims / 63 verified / 9 partial-body + 50 abstract-status all agree.",
    "Tally of 38 discriminating contrasts with no positive qualifying unit - consistent with the inventory; positives (AD Chlamydia culture, salmonid residency) are retained at their own restricted scopes rather than discarded."
  ],
  "residual_limitations": "Bounded access (9 partial-body, 50 abstract/status); six of the nine body-level sources were re-read here, three were not. Whether sequenced bacterial extraction blanks exist for PMID 29881346 remains unresolved in the accessed passages.",
  "divergent_copy_not_touched": "hsef_v53_popularity_ranks_31_35_2026-10-08/resident-brain-microbiome/resident-brain-microbiome.review.md - the earlier edition; it carries the same Momentum error, the same compressed PMID runs and the same access-based confidence rationale, but not the false-approval clause. Left untouched per instructions.",
  "approval": "Approved as revised."
}