Multi-source verifier · dry-run

Try the verifier on any question.

Paste a forecasting question, the resolution criterion you'd write for it, and the public sources you'd cite at resolution. The verifier returns a structured result — verifiable · pending · ambiguous · refused — with confidence and reasoning. It's allowed to say "I don't know with high confidence" when the criterion doesn't yield a determinate answer.

QuestionA yes/no question that can be answered definitively at resolution.

Resolution criterionThe exact rule the verifier will follow. Should name a concrete event + a comparison.

Named primary sourcesOne source per line. Must be public + cite-able.

Market closes atISO 8601 datetime.

Fill the form and run the dry-run to see the structured verifier output — status, confidence, reasoning, cited sources.

How the verifier decides

verifiable — criterion + sources unambiguously yield YES or NO at this point in time. Confidence ≥ 0.85 required; lower confidence auto-downgrades to ambiguous.
pending — market is still open OR the deadline has not passed yet OR the event has not happened yet.
ambiguous — criterion is real but the answer requires interpretation beyond what the criterion + sources alone can settle. Routes to the human appeal panel.
refused — criterion is malformed, distress-coded, or unverifiable in principle. The verifier refuses rather than guesses.

Same surface is exposed via the Foresight MCP server (foresight_resolve_check) so AI agents can call it without writing a UI.