diff --git a/doc/plans/2026-10-02-native-completion-calibrated-manifest.json b/doc/plans/2026-10-02-native-completion-calibrated-manifest.json index 6c24a8a0d0e..00ef3c5687c 100644 --- a/doc/plans/2026-10-02-native-completion-calibrated-manifest.json +++ b/doc/plans/2026-10-02-native-completion-calibrated-manifest.json @@ -267,12 +267,17 @@ "onlyDeclaredElevenNativeSourcesDiffer": true, "fixedMasterContextVerified": true, "providerCallsSoFar": null, - "liveStatus": "candidate_dispatched_baseline_approval_pending", + "liveStatus": "both_dispatched_no_graded_cells", "workflowRuns": { "candidate": { "id": 37098728980, "url": "https://github.com/paperclipai/paperclip/actions/runs/37098728980", "createdAt": "2026-10-03T05:06:09Z" + }, + "baseline": { + "id": 37098815696, + "url": "https://github.com/paperclipai/paperclip/actions/runs/37098815696", + "createdAt": "2026-10-03T05:07:44Z" } }, "preservedZeroProviderCohorts": [ diff --git a/doc/plans/2026-10-02-native-completion-master-qualification.md b/doc/plans/2026-10-02-native-completion-master-qualification.md index 2f20c76e640..d0e4b51b4aa 100644 --- a/doc/plans/2026-10-02-native-completion-master-qualification.md +++ b/doc/plans/2026-10-02-native-completion-master-qualification.md @@ -6,7 +6,7 @@ The [immutable calibrated manifest](2026-10-02-native-completion-calibrated-mani Only the previously reviewed eleven native implementation/unit-test files differ between variants. All five setup repair files are identical: a controlled mock receipt SHA, first-line framed provenance parsing plus positive/negative calibration, and matching immutable parent declarations. Production default/manual/shared-prompt/operational-skill bytes still match master `59c07ede`. The verifier's exact-source rejection and all behavioral oracles remain strict. -Both previous twelve-cell hosted cohorts stopped before providers and remain preserved below. [The twelve-retained-log replay](2026-10-02-native-completion-calibrated-admission/retained-frame-replay.json) validates exact hashes/framed proofs with no change to original failing verdicts. No model behavior reroll has occurred. [Candidate campaign 37098728980](https://github.com/paperclipai/paperclip/actions/runs/37098728980) is dispatched against immutable `d6e59e471`; the trusted workflow source is `215586d127e97c9301d86e769a39a15c13298ca2`. No graded cells are available yet. The historical dispatch is pending normal approval review: its initial request was rejected based on an earlier no-additional-runs instruction, and the parent task is submitting the same bounded request with the current direct human authorization. No duplicate candidate or baseline has been launched. Current campaign usage/cost is not yet available. This qualification does not measure fixed native prompt removal or general coding quality. +Both previous twelve-cell hosted cohorts stopped before providers and remain preserved below. [The twelve-retained-log replay](2026-10-02-native-completion-calibrated-admission/retained-frame-replay.json) validates exact hashes/framed proofs with no change to original failing verdicts. No model behavior reroll has occurred. [Candidate campaign 37098728980](https://github.com/paperclipai/paperclip/actions/runs/37098728980) is dispatched against immutable `d6e59e471`; the trusted workflow source is `215586d127e97c9301d86e769a39a15c13298ca2`. No graded cells are available yet. [Historical campaign 37098815696](https://github.com/paperclipai/paperclip/actions/runs/37098815696) is also dispatched once against immutable `e74ed61a6`, using the same trusted workflow revision. Initial automatic approval review rejections were resolved through the normal root approval path with direct current human authorization. No duplicate campaign or behavioral retry occurred. Both campaign builds now overlap. Current campaign usage/cost is not yet available. This qualification does not measure fixed native prompt removal or general coding quality. ## Preserved prior qualification stages diff --git a/doc/plans/2026-10-02-stock-harness-paperclip-checklist.md b/doc/plans/2026-10-02-stock-harness-paperclip-checklist.md index 40505715e79..957014ab515 100644 --- a/doc/plans/2026-10-02-stock-harness-paperclip-checklist.md +++ b/doc/plans/2026-10-02-stock-harness-paperclip-checklist.md @@ -508,3 +508,5 @@ PR #14961 remains draft and unmerged. Corrected standalone candidate 0a9c5a751 / The selected #14961 `0a9c5a7` / `00a761b` hosted cohort also stopped before providers. All twelve receipts prove zero calls; source isolation, lineage, hydration and selected native binary pass. One unit calibration omitted the source SHA from its mocked valid receipt, causing strict verification to reject it under hosted metadata. [The preserved twelve-receipt audit](2026-10-02-native-completion-qualified-hosted-setup.json) is a setup result; six behavioral pairs remain unmeasured. Only that test setup and the next exact parent-anchor declarations are being corrected. Selected #14961 now freezes `d6e59e471` / `e74ed61a6` after the approved five-file setup-only repair. Both exact refs pass actual launcher local and explicitly synthetic hosted prepare→verify (no credentials/providers): 132/127 selected TypeScript, 128 Node each, one local Rust calibration each; synthetic mode truthfully records no Rust execution. [The calibrated immutable manifest](2026-10-02-native-completion-calibrated-manifest.json) keeps exactly six pairs/twelve initial expected provider turns, strict oracles and single attempts. Prior twelve-cell zero-provider setup cohorts remain intact; model outcomes are pending. + +The first actual-provider #14961 comparison is dispatched once on separate immutable targets: [candidate37098728980](https://github.com/paperclipai/paperclip/actions/runs/37098728980) and [historical37098815696](https://github.com/paperclipai/paperclip/actions/runs/37098815696), same trusted workflow215586d. Builds overlap; zero graded pairs at dispatch, usage not yet available. Normal review resolved the publication/baseline approval provenance gates. No paid behavior retry or merge occurred.