FEN meta synthesis
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

FEN × Meta is a human-governed validation layer for AI-extracted knowledge. Muse Spark 1.3 reads source documents (including low-resource-language dictionary entries and EU regulation texts) and proposes structured claims, each with a verbatim quote, a separate model interpretation, uncertainty level, and cross-document contradictions. A deterministic, code-only gate then verifies every quote against its attributed source document, flags contradictions, and hard-blocks acceptance of anything unverified (UI + API 409). Only a recorded human decision can accept, dispute, or reject an item; every decision keeps its full provenance chain (document → quote → model proposal with prompt hash → human verdict). The model proposes, the gate checks, humans decide — no AI output becomes accepted knowledge on the model's word alone.