Stage D: an author-reported declared-stack replay#
Recorded outcome: MATRIX_MATCHED. The author records that a fresh-context agent received a sealed participant kit without the oracle, collected every cell once on Node 22, and that the separate oracle returned
REPRODUCED. The raw evidence is withheld, so a public reader can validate this record's consistency but cannot independently replay or verify that outcome.
The A–D evidence rail#
| A — untouched closure | B — coherent nested seed | C — opaque aliases | D — detached bundle |
|---|---|---|---|
| Author-reported PASS — exit 0, score 100; zero findings | Author-reported FAIL — exit 1, score 60; one critical closure violation | Author-reported INDETERMINATE — exit 2; one opacity refusal | Author-reported VALID — exit 0; integrity verified; authenticity not established |
The author reports that every A–D replicate pair was byte-identical. The recorded experimental delta was one declared-stack lockfile mutation in B.
What happened#
| Attempt | Runtime | Outcome | Results created |
|---|---|---|---|
| 1 | v20.20.2 | PREREQUISITE_FAILURE | no |
| 2 | v22.22.2 | MATRIX_MATCHED | yes |
Attempt 1 verified the archive and then failed closed because Node 20 did not satisfy the Node 22+ prerequisite. It was retained as a prerequisite failure. Attempt 2 ran the collector once, returned an unscored result. The author reports that all 347 returned-evidence checksums passed before the separate oracle scored it.
What this record reports#
- The separate oracle reported a match with the recorded A-D outcome matrix from one fresh-context collection.
- The author reports that all 347 returned-evidence checksums passed.
- The author reports that each A-D replicate pair was byte-identical.
What remains unestablished#
- external-human adoption.
- independent community replication.
- cross-platform portability.
- source origin authenticity.
- installed or deployed state.
- production behavior.
- ecosystem generality.
- security.
- performance.
- causal product efficacy.
The public independent-witness ledger currently records 0. This run does not contribute to that count. Oracle non-disclosure, the single invocation, and execution chronology are process attestations; this run has no signed transcript or append-only invocation log. Attempt 1 retains its structured summary, but not its original raw stderr.
Verify the public record#
From a clean checkout:
npm ci --ignore-scripts
npm run evidence:stage-d:verify
Download the machine-readable public record. The verifier returns
RECORD_VALID only when the exact author-reported matrix, attempt chronology, producing command,
hash-commitment shapes, and claim boundary are present. It does not verify the withheld evidence,
rerun the private subject, or make the process attestations cryptographic.
Integrity anchors#
| Withheld artifact commitment | Author-reported SHA-256 |
|---|---|
| BCE 0.5.0 package | 2002ca9a8ce3f95e96e9653720d2a7244baf090acf15bf4a0c8a1d45f2cd4b1d |
| Sealed participant kit | eb989d04809553ced959c342773a425e0894c21ed920a4c2c6f7ffdca2eb81b9 |
| Returned results manifest | 3e1152efbeda193fa8c36b52b3add854e150a47f852de5cd1fb98256e35e1071 |
| Returned checksum manifest | 56fd689340988fcd6c85e0fa542854de3b8fb7420e8a024a228c566aae25fa88 |
| Results archive | fd3ecf2ef019ed6aaaba659719191a5915baf02cad717e9362829e96c721a8d8 |
| Oracle verdict | 0a44ba01a101fceb6c5c0f38d07e44a209495847aebcc3919812b5806d874b20 |
The raw archives are not public because the sealed subject contains non-public source. Their hashes are author-reported commitments, not proof that a visitor can independently inspect or verify the withheld bytes.