Outcome, duration and value for a case or a book, read from filed court records.Outcome, duration and value, from the record
The Registry

Every model here earned its status.

Model claims are easy to inflate and hard to audit — so we keep a registry that shows production models next to the entries that never earned the label, and a changelog that records the corrections, not just the wins.

Series
Model Card MC-001
Version
1.0
Updated
9 October 2026
Status
Published
Authors
Criterica Intelligence Research
Corresponding author: Jack Cosentino, founder
Cleared individually
PRODUCTION STATUS
Each model passes temporal-holdout and calibration review on its own
Every attempt recorded
THE FULL REGISTRY
Models that failed stay on the record, labeled as failed
Judge-level
JUDICIAL ASSIGNMENT
A first-class model feature, not a footnote
Federal and state
COVERAGE
US states and federal circuits, plus Canada, the UK and Australia

One model = one jurisdiction × one case type × one prediction target. The entries that stay experimental, stubbed awaiting sufficient data, or recorded as failed are not overhead — they are the gate working. Status changes only through the same verification protocol that governs every published claim.

Methodology changelog

Prediction infrastructure that never corrects itself is not being measured. These are the material governance events on the fleet — including the ones where we found our own numbers wanting and withdrew them.

JUL 2026
GOVERNANCE
Production fleet re-cleared: 10,470 predictive models

After the retrain, the production fleet stands at 10,470 predictive models, each cleared individually through temporal-holdout and calibration review. National-scope models joined the fleet only where they outperformed the narrower models beside them.

JUL 2026
TRAINING
Federal civil models retrained on real-record labels

Federal civil models were retrained on labels taken directly from the court-recorded manner of termination, so every target describes what the record states: how a case ended and how long it took. Promotion now also requires a unique model signature, so one model is counted once. Every retrained model was re-cleared through the same gates before returning to production.

JUL 2026
GOVERNANCE
Below-gate models demoted from production

A fleet-wide temporal-holdout review identified a small set of production models performing below chance on the most recent evaluation window. All were moved back to experimental status pending label and evaluation audit, and the production count was reduced accordingly.

JUL 2026
CORRECTION
Fleet-wide calibration metric withdrawn for re-derivation

An internal audit found a previously computed fleet-level calibration summary had been evaluated on an unrepresentative sample. The figure was withdrawn from every surface rather than corrected in place, and per-model calibration review continues under the standard gates.

JUL 2026
DATA INTEGRITY
Court-record corpus deduplication verified

The primary corpus was deduplicated on a composite source key after cross-source identifier collisions were found, and the unique count was verified by a gated row-count check before the production data was swapped. The pre-deduplication file was archived, not deleted.

JUL 2026
GOVERNANCE
Models with unverifiable training provenance quarantined

A provenance review found a small model family whose training data could not be traced to verifiable source records. The family was reverted to experimental status and its previously published metrics were withdrawn.

APR 2026
TRAINING
Universal retrain on real data completed

Every production model was retrained exclusively on real filed records.

Verify it yourself
Registry inspection

Under NDA: walk the registry itself — model entries, training windows, holdout results, version history, status transitions. What this page says reconciles to that file.

Methodology walk-through

A working session with the team on pipeline, gates, and evaluation protocol. Bring your hardest statistician.

Audit on your book

The strongest test: we score your historical matters with known outcomes and you measure the models against what actually happened. Your data, your baseline.

Arrange a verification session →
Cite this document

Criterica Intelligence Research. (2026). The Model Registry and Public Changelog (Model Card MC-001, version 1.0). Criterica. https://critericaintelligence.com/registry

@misc{criterica_mc001_2026,
  author = {{Criterica Intelligence Research}},
  title = {The Model Registry and Public Changelog},
  publisher = {Criterica},
  howpublished = {Model Card},
  number = {MC-001},
  year = {2026},
  month = oct,
  note = {Version 1.0},
  url = {https://critericaintelligence.com/registry}
}
Download PDF

DOI: pending

Next step

See what your book is worth before you commit capital.

Describe the book. After qualification you receive a written report, a per-matter schedule and a review session.

Request an assessment

Need funding for a case, a firm or a book? Criterica Capital