The registry — catch history

Every engagement, on the record. Including the misses.

An append-only, hash-chained log of verification outcomes, clinical-trial style: an engagement is logged when it opens, so a missing outcome is itself visible. Entries are redacted by construction — claim, metric, claimed vs recomputed, verdict, and content hashes; never code, never data. Each entry embeds the SHA-256 of the previous one and is signed with the lab key; the chain audits offline with one command, or any single entry with stock OpenSSH.

audit: python3 scripts/calma.py registry verify registry/

Don't take this page's word for it: the raw entries, the lab's public key, and the audit command above let anyone re-verify the whole chain offline — or any single entry with stock ssh-keygen.

The chain1 entry · head dc236f5759bb

No client engagements have been logged yet — the chain below opens with the lab's own demonstration catch. From the first engagement on, every outcome — confirmed, refuted, withdrawn — appends here permanently. A registry that only showed wins would be worthless; this one can't.

#000012026-06-11verificationself-test · demonstrationREFUTED

claimed total_return 146.97697947938846

claimed 147 (+14,697.7%) → recomputed -0.3243 (-32.4%)

dc236f5759bb4acd · genesis

Entry format: in-toto/DSSE attestation bundle → redacted whitelist → hash chain → SSHSIG. Verdicts can additionally be Sigstore-countersigned and witnessed by the public Rekor transparency log, offered per engagement. v2 adds a Merkle tree per the C2SP tlog-tiles spec — additive, the entries are already hash-addressed.

Questions about an entry, or a number you need verified? How the lab works