# Main disposition of the completed correction diagnostics

2026-09-07. **ACCEPT VERIFIED DIAGNOSTICS; PROCEED TO EDITORIAL/RELEASE
ASSEMBLY WITH QUALIFICATIONS.** This closes the one prescribed input/results
verification. It does not activate a release, alter frozen v6, or publish data.

Main read Peirce's complete REVIEW and FINDINGS, inspected the saved matrix,
comparison, direct-estimator and preservation checks, and rechecked all seven
review-artifact hashes and all763 unique inspected file bindings. All match;
both phase reports have empty error lists. No material finding or correction
is pending. This is not a second full numerical rerun or a renewed source audit.

## Accepted evidence and scope

- Frozen675 becomes a private676-row candidate through13 retirements and14
  separately qualified full rows. The1,876 compelled retained cells comprise
  **927 changed and949 confirmed**;37,844 untouched cells preserve their exact
  strings and U/NA types, including100 inherited fractional values.
- Actual run `model_execution_v1/run_20260907a` contains565 labeled scenario
  records,536 distinct preprocessed matrix cases and538 checkpoint units.
  Repeated/related scenarios are not independent experiments.
- The reviewer independently reconstructed all565 matrices and reported
  comparison calculations, and directly reproduced six selected factor fits
  and four500-draw Horn cases. It did not reproduce every loading fit/null
  simulation. Its prior upstream work, including Circeii assembly, is disclosed;
  this is a model-performed check independent of main's present assembly,
  not a human/external audit or historical revalidation of its own work.

## Required interpretation

1. Both canonical datasets retain nine factors, but the fitted relationships
   change. Candidate versus baseline maximum angles are18.689067 degrees in
   forced-eight space and11.784825 degrees in forced-nine space. Minimum named
   axis Tucker congruence is0.077522. The named-axis comparison retains its
   labels; it is not the separate optimally rematched whole-space comparison.
2. Legacy k6 reassignment is17.975831% across662 shared rows (ARI0.632360).
   This is a diagnostic of an old automatic partition, **not** a vote to
   reassign17.98% of the six reviewed editorial families.
3. The record distribution is524 Horn9,29 Horn8 and12 Horn7. Separate
   disciplinary-battery omissions retain6–8. Literal linked-program weighting
   retains676 actual rows, total weight583 and effective sample size615.777;
   it changes forced-eight space by44.671536 degrees relative to the candidate.
   Keep these results visible. There is no prespecified universal pass/fail
   angle, and no rule to choose historical boundaries to improve stability.
4. The eight public trait cards remain fixed-anchor descriptive summaries,
   not newly estimated latent-factor scores. Recalculate them from corrected
   raw cells using the unchanged card model, preserving unknown/NA and bounds.
   Carry forward only explicit accepted editorial family choices and holds.
5. Date alternatives are nominal-chronology diagnostics, not measured movement
   durations. Calendar envelopes, historical tempo indicators and observation
   horizons remain distinct. Egypt-through642-only and internal Kitakami
   phase profiles remain explicitly unestimable; no pseudo-scores fill them.
6. These diagnostics demonstrate the consequences of specified boundaries,
   missingness and instrument choices. They do not prove an unbiased census,
   historical validity by numerical stability, universal traits, or a complete
   release-wide fairness conclusion beyond the disclosed tests and limits.

## Remaining implementation, not another research gate

Build the private676-row descriptive cards and exact-ID family projection;
test inheritance, holds, coverage and output bindings. Record any unresolved
editorial alternative as such, not an inferred classification. Complete the
approved publication-bound text checks on the changed results/claims, prepare
the immutable successor bundle, then synchronize release labels and public
artifacts through the authorized deployment workflow. These are downstream
obligations, not a reason to rerun this closed verification.

`MAIN_DIAGNOSTIC_DISPOSITION.json` binds this decision to the immutable run
and reviewer findings. GitHub remains private and Zenodo unpublished; neither
publication switch is authorized by this disposition.

## Same-day card-integration clarification

The instruction in item4 to use the “unchanged card model” was underspecified:
the frozen v6 model still has the rejected five-anchor F8, while the public
method already adopted direct, applicability-gated **Imported unfree labor**
in August. Preserve the unchanged-v6 calculation as a historical comparison;
use the already approved F8 definition in the separate current-public card
layer, with F1–F7 unchanged. See
`editorial_projection_v1/MAIN_F8_INHERITANCE.md`. Its eight adjudications and
three display corrections do not themselves close six earlier raw-coding
follow-ups. This clarification does not change the accepted numerical run,
its bindings, raw inputs or the findings disposition.