Issues

• moments: other "siding with the opponent" spellings still read as a side's own merits filing

  ModelMirrorAI/fedcourtsai#2370 · Filter by author modelmirror opened on Oct 11, 2026

• governance: a separate, non-admin GitHub identity for interactive agent sessions, so review rules bind them

  ModelMirrorAI/fedcourtsai#2365 · Filter by author modelmirror opened on Oct 11, 2026

• promotion: a prep workflow that syncs, runs the integration test and pre-flight, and refreshes the promotion PR, so the maintainer only merges

  ModelMirrorAI/fedcourtsai#2363 · Filter by author modelmirror opened on Oct 11, 2026

• dev: a maintainer-gated, read-only corpus check that runs a PR's declared check and posts the result

  ModelMirrorAI/fedcourtsai#2364 · Filter by author modelmirror opened on Oct 11, 2026

• governance: enforce the maintainer-only change classes with CODEOWNERS and a code-owner ruleset on staging

  ModelMirrorAI/fedcourtsai#2361 · Filter by author modelmirror opened on Oct 11, 2026

• orchestration: chain pull → predict → evaluate on run completion, and drain backlogs under a pre-registered spend cap

  ModelMirrorAI/fedcourtsai#2362 · Filter by author modelmirror opened on Oct 11, 2026

• release: a committed release manifest the site reads, so publishing a release needs no fedcourts.ai change

  ModelMirrorAI/fedcourtsai#2356 · Filter by author modelmirror opened on Oct 11, 2026

• process: continuous promotion — windows open at the carrying merge automatically, and results pool valid predictions across protocols by default

  ModelMirrorAI/fedcourtsai#2346 · Filter by author modelmirror opened on Oct 10, 2026
  3

• salience: an interim reserve slot stays occupied while its application's disposal goes undetected

  ModelMirrorAI/fedcourtsai#2340 · Filter by author modelmirror opened on Oct 10, 2026

• predict: a granted side motion reads as an interim disposal and parks a pending application's forward cells

  ModelMirrorAI/fedcourtsai#2339 · Filter by author modelmirror opened on Oct 10, 2026

• runner: GeminiRunner (the back-test's real-engine path) does not retry a zero-exit INVALID_STREAM

  ModelMirrorAI/fedcourtsai#2321 · Filter by author modelmirror opened on Oct 9, 2026 · CI & engine hygiene

• corpus: maintenance commands' default dry runs open the corpus read-only, so they stop migrating an older pulled blob in place

  ModelMirrorAI/fedcourtsai#2319 · Filter by author modelmirror opened on Oct 9, 2026 · CI & engine hygiene

• corpus: an explicit remote-writes switch set only by the writer jobs, so a non-writer Actions job never attempts a mirror write

  ModelMirrorAI/fedcourtsai#2318 · Filter by author modelmirror opened on Oct 9, 2026 · CI & engine hygiene

• ci: move the pinned runner image from ubuntu-24.04 to ubuntu-26.04 after the long-conference release

  ModelMirrorAI/fedcourtsai#2310 · Filter by author modelmirror opened on Oct 8, 2026

• salience/scope: a hold-candidate signal — petitions held for a merits decision (GVRs) are selected only by accident, and IFP ones never

  ModelMirrorAI/fedcourtsai#2299 · Filter by author modelmirror opened on Oct 8, 2026

• predict: a cell can restore the removed qp-topics oracle from git and read other predictors' committed output

  ModelMirrorAI/fedcourtsai#2289 · Filter by author modelmirror opened on Oct 7, 2026
  1

• retrieval: gemini's retrieval log records no call results (and codex code-mode calls none), so quota refusals cannot be measured per engine

  ModelMirrorAI/fedcourtsai#2278 · Filter by author modelmirror opened on Oct 6, 2026
  1

• grading: stamp the cert skill anchor harness-side and pin the statpack build per cohort (proc-v9)

  ModelMirrorAI/fedcourtsai#2272 · Filter by author modelmirror opened on Oct 6, 2026
  3

• documents: the contact scrub never runs when the docket serves no petitioner block (pseudonymous self-represented filer)

  ModelMirrorAI/fedcourtsai#2264 · Filter by author modelmirror opened on Oct 6, 2026

• predict: a cell cannot look up a companion docket by number, and a Vide-linked docket's state is never staged

  ModelMirrorAI/fedcourtsai#2263 · Filter by author modelmirror opened on Oct 6, 2026

• blinding: record what un-aliasing replaced, so a judge-written predictor name is distinguishable from a harness substitution

  ModelMirrorAI/fedcourtsai#2262 · Filter by author modelmirror opened on Oct 6, 2026

• metrics: the cert skill anchor is judge-transcribed, and the refresh body misstates n, ties and coverage at scale

  ModelMirrorAI/fedcourtsai#2245 · Filter by author modelmirror opened on Oct 5, 2026
  3

• interim: forward cells run on or after the day an application is decided (resolved_at has no time of day)

  ModelMirrorAI/fedcourtsai#2243 · Filter by author modelmirror opened on Oct 5, 2026
  1

• predict: first-time distribution-moment cells are minted after the petition's conference has passed

  ModelMirrorAI/fedcourtsai#2241 · Filter by author modelmirror opened on Oct 5, 2026
  3

• backtest: decide whether party dispositions (dismissed/withdrawn) belong in the cert back-test set

  ModelMirrorAI/fedcourtsai#2217 · Filter by author modelmirror opened on Oct 3, 2026 · Back-test program