Bundle 70: Fresh Personal Holdout Runner R1

[nulleimy]

Context

Canonical base: feature/bundle-0-bootstrap @ 27af5d04e1f7971126adef74dff4d34ee54536df

Bundle 69 made the R2/R3 methodology valid, but there is no local execution runner that turns a private local-library snapshot into a fresh preregistered personal holdout and reviewer-safe workspace.

Goal

Add a local-only runner that creates a fresh personal blind curation holdout with >=24 effective cases, four per set role, frozen fallback policy, frozen challenger evidence before review, and a reviewer-safe sequence-only packet.

Required invariants

• no human labels or challenger scores may influence holdout selection;
• challenger evidence must be frozen after selection but before reviewer workspace publication;
• reviewer packet exposes no optimizer strategy, shadow score, challenger preference, or absolute path;
• only four R2 curation dimensions are requested;
• transition execution is not requested;
• failed candidate cases must be isolated as technical invalidities rather than aborting the full pool;
• no fabricated ratings/preferences/confidence/timestamps;
• no optimizer/release/deploy/production/PDM authority.

Authority

MERGE_AUTHORIZATION=NO
OPTIMIZER_RANKING_ACTIVATION=NO
PDM_TRAINING=NO
RELEASE_AUTHORIZATION=NO
DEPLOY_AUTHORIZATION=NO
PRODUCTION_ACTIVATION=NO

[nulleimy]

Bundle 70 final Ready-for-Review receipt

Exact HEAD: e3c8d3e27e6b993527c6bd9c39a14bcd0cbca890

Verification:

• CI #1020: SUCCESS
• Python 3.11.16: 531 passed, 6 warnings; artifact SHA-256 61095b4bfd11a8cfa25d6b49551ebaba67b75469071805121e5ba0796c7d3e80
• Python 3.12.14: 531 passed, 5 warnings; artifact SHA-256 32663e93eb84eb347f8e95b143bf7320c0f72b1280449c40f7dfee429d4a1c96
• compile + critical Ruff PASS on both
• PR Guard #377, #378, #379: SUCCESS
• scope: 9 new Bundle70 files, 0 existing files modified/deleted, behind canonical 0
• optimizer authority unchanged; Set Engine / Set Path Optimizer / Bundle67 Competitive Curation blobs unchanged
• final Skill Tester: PASS; no unresolved HIGH/CRITICAL findings

Hardening closed: per-case technical failure isolation; exact canonical clean-checkout preflight; challenger freeze before reviewer publication; pre-finalization byte-hash binding; exact effective-cohort/assignment/role binding; empty human/attestation fields at freeze; dual prior-exposure exclusion using reviewer-visible sequence + stable private track-ID evidence; reviewer leakage protection; zero replacement events before publication; immutable R1 cohort during review.

PR #166 is Ready for Review. Actual Fresh Personal Holdout Run remains blocked until Bundle70 is canonical.

MERGE_AUTHORIZATION=NO
OPTIMIZER_RANKING_ACTIVATION=NO
PDM_TRAINING=NO
RELEASE_AUTHORIZATION=NO
DEPLOY_AUTHORIZATION=NO
PRODUCTION_ACTIVATION=NO

Merge requires separate explicit MERGE GO.

[nulleimy]

Bundle 70 pre-merge receipt — late review remediation complete

User merge authorization: explicit MERGE GO.

Exact feature HEAD: baa7e8312ef02907e01042ab4fb6d841720e5108
Canonical base: 27af5d04e1f7971126adef74dff4d34ee54536df

Verification:

• CI #1026: SUCCESS
• Python 3.11.16: 533 passed, 6 warnings; artifact ID 9492270480; SHA-256 44adee1ea241e9b2620c3059a5761eaeab8788c2de40a7214c41ce0285bb4cb0
• Python 3.12.14: 533 passed, 5 warnings; artifact ID 9492269541; SHA-256 65d8089109b87b6b700a892a8790cf21558ec0b4be26bf504ec6607d852c57cd
• compile + critical Ruff PASS on both
• PR Guard #382 and post-description Guard #383: SUCCESS
• branch behind canonical: 0
• scope: 9 added Bundle70 files, 0 existing files modified/deleted
• Set Engine blob unchanged dd8f0cdb672202289530579d092cf0feb5cedf25
• Set Path Optimizer blob unchanged 819f6399181f4b37bb26ed8ccefb6cc0f4a9607f
• Bundle67 Competitive Curation blob unchanged faadcae932e5f7b907503cbf2cd5bfdaab36462e

Late Codex review surfaced two P1 findings after the first Ready gate. Both are fixed on the exact HEAD above, covered by regression tests, replied to with evidence, and their review threads are resolved:

1. MIR/analyzer failures are isolated at the per-case boundary instead of aborting shared candidate evidence generation.
2. Full source CuratedReviewCase records are frozen privately before reviewer publication so later unblinding/R3 calibration does not require optimizer rerun.

MERGE_AUTHORIZATION=YES (explicit user MERGE GO)
OPTIMIZER_RANKING_ACTIVATION=NO
PDM_TRAINING=NO
RELEASE_AUTHORIZATION=NO
DEPLOY_AUTHORIZATION=NO
PRODUCTION_ACTIVATION=NO

[nulleimy]

Bundle 70 post-merge receipt

PR #166 merged successfully.
Feature HEAD: baa7e83
Canonical merge SHA: 375a8d7

Post-merge checks:

• Issue Bundle 70: Fresh Personal Holdout Runner R1 #165 closed as completed.
• feature/bundle-0-bootstrap is identical to the merge SHA (ahead 0, behind 0).
• set_engine.py blob remains dd8f0cdb672202289530579d092cf0feb5cedf25.
• set_path_optimizer.py blob remains 819f6399181f4b37bb26ed8ccefb6cc0f4a9607f.
• competitive_curation.py blob remains faadcae932e5f7b907503cbf2cd5bfdaab36462e.
• CI #1026 passed with 533 tests on Python 3.11 and 533 tests on Python 3.12.
• PR Guard #382 and #383 passed.
• Both late P1 review findings were fixed and their threads resolved before merge.

Only the requested PR merge was performed.