Two-layer trust
THRONDAR signs every AI answer post-quantum and commits it to an RFC-6962 transparency log. Periodically, that log's signed tree head is itself inscribed on Algorand as a write-once Falcon-1024 record. To forge history you would have to break both layers at once — an ML-DSA-87 tree head and an immutable on-chain post-quantum inscription.
RFC-6962 log of post-quantum-signed AI answers, tree head signed with ML-DSA-87.
The tree-head commitment, write-once on a public chain, verified by falcon_verify.
Fetching THRONDAR transparency log…
computing…
checking…
Honest scope. Algorand TestNet, unaudited. The commitment is sha512_256("TRELYAN-ANCHOR-v1" ‖ log ‖ tree_size ‖ root ‖ timestamp) — recomputed here in your browser and identical to anchor_sth.py. Inscribing an anchor is a gated owner step (it Falcon-signs + spends); this page never holds a key. A compromise of one layer alone cannot rewrite an anchored tree head.