Issues

• backtest: a merits replay arm, and the merits history it needs (OT2025 argued cases first)

  ModelMirrorAI/fedcourtsai#2602 · Filter by author modelmirror-agent opened on Oct 21, 2026

• scoring: review how non-standard dispositions are scored — GVRs, summary reversals, DIGs, and merits cases that end without a decision

  ModelMirrorAI/fedcourtsai#2601 · Filter by author modelmirror-agent opened on Oct 21, 2026

• documents: a cell can run before the freshness pass fetches a filing already on its snapshot

  ModelMirrorAI/fedcourtsai#2600 · Filter by author modelmirror-agent opened on Oct 21, 2026

• documents: Rule 34.6 dockets reach cells with no filings, and the cell is not told why

  ModelMirrorAI/fedcourtsai#2599 · Filter by author modelmirror-agent opened on Oct 21, 2026

• live rotation: ~25k OT2017–OT2024 cert-form rows are rotation-eligible but never polled

  ModelMirrorAI/fedcourtsai#2581 · Filter by author modelmirror-agent opened on Oct 19, 2026

• predict prompt: the documents item lists only four kinds and two manifest fields

  ModelMirrorAI/fedcourtsai#2579 · Filter by author modelmirror-agent opened on Oct 19, 2026 · Next re-bless: proc-v10 / sal-v5 candidates

• documents: size the multi-respondent oppositions whose briefs are all scans

  ModelMirrorAI/fedcourtsai#2578 · Filter by author modelmirror-agent opened on Oct 19, 2026 · Provisioning & document coverage

• documents: re-measure text coverage; the published empty-text figures date from the 2026-08-28 blob

  ModelMirrorAI/fedcourtsai#2577 · Filter by author modelmirror-agent opened on Oct 19, 2026 · Provisioning & document coverage
  1

• Weekly performance digest — 2026-W43

  weekly-digest
  ModelMirrorAI/fedcourtsai#2555 · Filter by author github-actions opened on Oct 19, 2026

• promotion train: wake on the gate run finishing (and on approval) instead of waiting for the hourly tick

  ModelMirrorAI/fedcourtsai#2542 · Filter by author modelmirror-agent opened on Oct 18, 2026

• corpus-check: corpus-info --text-coverage exits 1 inside the lane; surface why, then fix

  ModelMirrorAI/fedcourtsai#2533 · Filter by author modelmirror-agent opened on Oct 18, 2026

• backtest: read replay cells' flags.json through the normaliser, so a mis-shaped file isn't recorded as unreadable

  ModelMirrorAI/fedcourtsai#2515 · Filter by author modelmirror-agent opened on Oct 17, 2026 · Back-test program

• outcome: a re-firing mint rewrites event.yaml's caption, so the re-docket check loses its committed witness

  ModelMirrorAI/fedcourtsai#2514 · Filter by author modelmirror-agent opened on Oct 17, 2026 · Waiting on data or a decision
  1

• Automate the promotion's sync → test → prepare sequence; make the contexts check on-demand

  ModelMirrorAI/fedcourtsai#2502 · Filter by author modelmirror-agent opened on Oct 17, 2026

• metrics: pre-register a self-interest comparison — predictors forecasting litigation against their own developer

  ModelMirrorAI/fedcourtsai#2458 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Lower-court slices: AI cases and the D.C. Circuit

• resolution: deterministic district ruling resolver for the watchlist's trial moments

  ModelMirrorAI/fedcourtsai#2457 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Lower-court slices: AI cases and the D.C. Circuit

• district: pre-register a court-agnostic trial stage — moments and outcomes for motions, judgment and settlement

  ModelMirrorAI/fedcourtsai#2456 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Lower-court slices: AI cases and the D.C. Circuit

• corpus: ingest the AI watchlist's dockets, district courts included, without admitting their courts

  ModelMirrorAI/fedcourtsai#2455 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Lower-court slices: AI cases and the D.C. Circuit

• ai-cases: commit the AI-case watchlist and its rule, and tag its cases across courts

  ModelMirrorAI/fedcourtsai#2454 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Lower-court slices: AI cases and the D.C. Circuit

• ai-cases: AI-case watchlist slice — tracking (milestone 2 in docs/milestones.md)

  ModelMirrorAI/fedcourtsai#2453 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Lower-court slices: AI cases and the D.C. Circuit

• feeds: an AI-cases feed — just the facts on watchlist dockets, predictions as opt-in enrichment

  ModelMirrorAI/fedcourtsai#2452 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Public update feeds

• feeds: optional public email updates — weekly SCOTUS update and AI-cases feed — tracking

  ModelMirrorAI/fedcourtsai#2449 · Filter by author modelmirror-agent opened on Oct 15, 2026 · Public update feeds

• feedback: make agent flags triageable — never fail a cell, collapse re-reports, mark tracked themes (harness now; prompt in proc-v10)

  ModelMirrorAI/fedcourtsai#2442 · Filter by author modelmirror opened on Oct 15, 2026 · Agent observability

• predict: gemini-baseline's retrieval.md says no retrieval while its harness log records corpus queries

  ModelMirrorAI/fedcourtsai#2425 · Filter by author modelmirror opened on Oct 14, 2026 · Agent observability
  2

• predict prompt: a cert "distribution" cell reads the latest snapshot uncut, but predict.md says it is forecast at first distribution

  ModelMirrorAI/fedcourtsai#2423 · Filter by author modelmirror opened on Oct 14, 2026 · Next re-bless: proc-v10 / sal-v5 candidates