Oc2cO

ClaimWatch is a self-verifying research agent: a verification layer that wraps any research or digest pipeline and answers two questions about every run — did every leg actually land, and is what it says actually true? It emits per-leg receipts, checks each promised artifact, and fails empty runs outright instead of reporting success. It extracts checkable claims, re-fetches cited sources, runs independent sanity checks, and classifies each claim as Verified (substance), Unverified, Refuted, or Pending, with evidence attached. Every decision lands in an append-only SHA-256-chained ledger anyone can re-verify, with replay verification and an offline dashboard. Built with Muse Spark via the Meta Model API for the verification reasoning. The included example runs are mock/replay; a separate archived proof covers one fictional synthetic live case. All demo data is synthetic and sanitized.