MUSE Evidence Review Core

MUSE Evidence Review Core is an open-source editorial workspace that helps writers identify unsupported claims without surrendering editorial control to AI. The application separates the original source, a human-saved draft, and an AI evidence review. Using Meta Muse Wonder 1.0, it evaluates whether claims are supported, contradicted, not established by the source, or require human interpretation. MUSE independently verifies exact source quotations and binds each review to a specific source and saved draft revision. Its defining feature is STALE invalidation: when the writer modifies and saves the draft, the previous review immediately becomes outdated and cannot be treated as approval of the new version. The live demo demonstrates real Meta inference, source quotation verification, and human-controlled revision. Unlike a basic RAG chatbot, MUSE combines evidence traceability, revision integrity, and explicit human decision-making. The standalone competition core was newly implemented during the hackathon, informed by earlier private research. No private implementation is included or required.