Cryptographic provenance & implementation evidence ledger

The code passed; no accepted performance corpus exists yet

This evidence ledger publishes the immutable, machine-readable provenance records for the benchmark suite. It links git commit hashes, SHA-256 source file digests, toolchain versions, and compiled binary artifacts to prevent drift and ensure independent inspectability.

Accepted performance corpus: none. M1 starts checking precision after 20 committed pairs in each stratum. A stratum can also finish inconclusive after its 60th planned attempt. The removed manual validation records cannot supply those pairs.

Open acceptance JSON Multi-language kernel provenance Base catalog implementation status Audio v2 proposal records

Inspect the exact source and build

The panel binds each file and SHA-256 to the accepted commit. Only the fixed build-manifest and compiled-Wasm routes are available as local downloads.

Open the source/build manifest without JavaScript

Loading the enhanced source and build panel…

What was accepted

Source commit: 9c309c4941d1b8550c15f8549f95a5636a634ef6
Full gate: deno task check: 16 passed, 0 failed, 0 skipped; 15 negative contract invariants
Wasm artifact: 96 bytes · SHA-256 9c4ce5f0d9e32cdd364b73b2697566e7396368d9867d9bc3d939bb2063583a6d
Independent review: ACCEPT · run f2f58de5 · no blocker, high, or medium finding

Chrome 150 attestation: unverified

A parent attestation records four observations:

The visible “Validate and run pair” control completed for JavaScript and linear Wasm.
WebAssembly compiled under the narrow wasm-unsafe-eval CSP allowance; general eval and inline execution stayed disabled.
Both variants passed correctness and fixed-work validation, and two uploads returned HTTP 201.
The owned loopback server stopped, the two validation records were removed, and the MCP page closed.

No screenshot, run record, or replayable MCP log was retained. Independent review therefore cannot verify these observations. The acceptance JSON records that limit and the reported cleanup steps.

Evidence still required

The project has no published raw validation sample, cold/warm fresh-profile corpus, paired ratio, exact interval, stopping decision, or performance conclusion. M1 performance acceptance requires the preregistered fresh launches through the precision target or attempt cap.