Start here
- Final Freeze Note Plain-language final lock document. What the corporate extension is, what it is not, the geometry roles, and the operator’s claim discipline.
- Phase 19 Decision Note Integrated two-label decision across the five Phase-19 hardening tracks. The honest "signal survives but is weaker than simple baselines" framing lives here.
Core validation
- Phase 18 Full-Population Results The locked Phase-18 final result on the full eligible-universe panel. Pooled and per-year metrics; the comparison chain across Phases 15–18.
- Baseline Comparison Head-to-head: A/G composite vs single-primitive baselines vs an Altman-style proxy. The reviewer-defense write-up of the "single primitives beat the composite" finding.
- Sector-Stratified Analysis Per-SIC-division metrics with explicit reliability flags. The signal is broadly distributed but Manufacturing-3 (durables / electronics) is the weak cell.
- Calibration Policy Predeclared monitoring-band policy for the structural-distress score. Watch (80–95th percentile) is the most informative tier; the top-1% Critical band shows a tail dropoff.
- Divergence-Anchor Study Internal future-state anchor study. First internal evidence that the canonical R / divergence geometry behaves as predicted for its intended regime.
- Anchor Refresh Note A bounded EDGAR FTS refresh from 2026-01-01 onward; 6 new Item-1.03 anchors and a 22-row out-of-time descriptive check.
Reproducibility & artifacts
-
Artifact Index
A complete map of current vs historical vs superseded artifacts (markdown notes, scripts, processed CSVs) lives in the project repository as
CORPORATE_ARTIFACT_INDEX.md. Not duplicated on this public site to keep the surface focused. -
Evidence Packet Manifest
A compact send-back manifest for reviewers — what to include in a packet, what to exclude — lives in the project repository as
CORPORATE_EVIDENCE_PACKET_MANIFEST.md. -
Reproducibility Runbook
Step-by-step instructions to re-derive the locked Phase-18 / Phase-19 outputs from raw SEC data live in the project repository as
CORPORATE_REPRODUCIBILITY_RUNBOOK.md.
Why this site does not distribute data
The Phase-18 firm-year panel and per-firm-year scores are not distributed from this site. Reasons:
- The panel is large (~30 MB) and contains per-CIK firm-year fields that could be misread as a public ranking.
- Publishing per-firm distress scores risks being repurposed as a "watchlist" — explicitly outside what the corporate extension claims.
- Anyone wanting to reproduce can re-fetch from the public SEC EDGAR endpoints using the published Phase-18 / Phase-19 scripts in the project repository.