Open fixtures · independent implementations · explicit status
Bring your
hardest fixture.
Run the same CBOR‑LD document through Swift, JavaScript, Python, and Rust. See every cross-decode, separate semantic agreement from exact bytes, and inspect where commercial acceleration has—and has not—been measured.
Participants
One corpus. Equal rows.
No implementation receives a preferred correctness rule. Every active encoder is crossed through every active decoder for the selected fixture.
Raw CBOR oracle— selected vectors checked independently; legacy rejection remains explicit.
Schema and alternate implementationsCDDL and Java resources remain source-present until their executable lanes run.
Cross-decoding explorer
Every encoder meets every decoder.
Select a fixture to inspect its complete decode grid, emitted bytes, and exact-consensus status.
| Loading cross-decodes… |
Release-mode measurements
Inspect the operation, not only the headline.
In-process measurements load from the latest report.
Commercial acceleration · evidence held open
SemanticCompute gets a track, not a free pass.
Catalogue presence, adapter source, binary execution, parity, and performance are different claims. The lab will not convert one into another.
SemanticCompute is not required to build, test, or use Swift‑CBORLD core. Any future accelerated result must identify the SC version, binary hash, backend, hardware, fallback status, and public CPU reference.
Community challenge lane
Make the corpus harder.
Contribute a valid edge case, a malformed specimen, a context rule, a canonical-byte disagreement, or a scaling document. The value is a portable regression fixture—not a promotional issue.