Proof-of-claim receiptsas trust infrastructure.
Verified is built on a narrow idea: when an agent or operator is challenged, it should be able to produce a proof-of-claim receipt showing exactly what was checked and what the system concluded.
1.The trust problem is not that AI answers exist. It is that a confident output does not show what evidence was used, whether independent models disagreed, or whether the record changed later.
2.The protocol thesis is independent verification infrastructure: one factual claim, supplied evidence, one primary route and three heterogeneous judges, then deterministic adjudication and a signed receipt.
3.The canonical transaction is narrow on purpose. It canonicalises the submission, screens it through the firewall, routes the same evidence to the judges, and returns one of three public signals: VERIFIED, CHALLENGED or UNVERIFIED.
4.Receipts should be inspectable, shareable and immutable. Verification is only useful if the receipt can be audited long after the event.
5.Verified does not solve for ultimate truth. It solves for structured, evidence-based artifacts that let the customer prove which process ran, under which policy, against which evidence, at which time.