All pull requests

Pull requests

• TE-F3c: eval check on question-level results (exact stratified permutation rule; replaces F3-b)

  #584·abhishek255 opened Nov 14, 2026Contributor·

  ·

• feat: pass approval policy through decision service

  #582·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Task actions 3/3: continue, more work for a task that ended its run (DECISIONS 198)

  #581·abhishek255 opened Nov 13, 2026Contributor·

  ·
  1

• Task actions 2/3: review findings, accept with a reason or fix each (DECISIONS 198)

  #580·abhishek255 opened Nov 13, 2026Contributor·

  ·
  1

• Stage selection: small changes skip planning, with the reviewer's questions and the grow-after-build stop

  #579·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Task actions 1/3: send back to coding (DECISIONS 198)

  #578·abhishek255 opened Nov 13, 2026Contributor·

  ·
  1

• refactor: export host-neutral session actions

  #577·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Report file entry types and discover instruction files

  #576·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-C5: additive machine sizing and Prove measurements

  #575·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Repin AWS Emulator service table to 1.0.343

  #573·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Change tier classifier: estimate, diff tier, merge recompute, catch counters

  #570·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-C1: baseline cached per base commit (nearest trusted ancestor, trusted writers only)

  #561·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-C6: rerun a failing test once in a task round; flaky is not a regression

  #559·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Shard the approval-proof markers over 32 partitions, still honouring the old key

  #556·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-A6 Part A: pure per-test attribution join (no callers)

  #554·abhishek255 opened Nov 13, 2026Contributor·

  ·

• Keep Google authentication hosts out of model blocks (B+)

  #550·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F3-b: eval gating (DECISIONS 170/171): paired noise band, holdout limits, steady eval, measures v2

  #549·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-B11: add disposable mutation spot checks

  #548·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F1c: model bridge, real modes, ledger, phase A cache use, laptop adapter (draft, do not merge before TS-L1-L3)

  #547·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F11: add reply-health checker and expiring probe store

  #545·abhishek255 opened Nov 13, 2026Contributor·

  ·

• [TE-D1] Add policy resolution and the PolicySource read seam

  #544·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F3: eval check (verdict and noise band, base score cache, pacing, hidden set, worker round)

  #543·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F10: side-by-side old vs new prompt outputs with a pinned independent judge

  #541·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F8: record-then-replay for changed prompts (also carries F1-a b794a4f7d)

  #540·abhishek255 opened Nov 13, 2026Contributor·

  ·

• TE-F1b: model helper network and by-name model-host block (Code half, draft, merge after TS-L1-L3)

  #539·abhishek255 opened Nov 13, 2026Contributor·

  ·