RERUN

Checked twice.

A forecasting market on whether published research holds up. People predict the result, then the Lab, an AI agent, reruns the result from public code and data and posts a receipt on NEAR that anyone can check.

Other truth markets trust an AI's opinion. We rerun the research and post the receipt.

Live on NEAR mainnet: https://rerun-mainnet.vercel.app (real USDC) · Testnet, free points: https://rerun-testnet.vercel.app · Video: https://x.com/Maveriq_/status/2121174144926847265

Try it in 2 minutes

No setup and nothing to pay:

• See a verdict, no wallet: open claim 5 on mainnet, then its receipt. Your browser re-checks the receipt against what's stored on NEAR.
• Forecast for free: on the testnet site, connect a testnet wallet (Meteor or Intear make a free testnet account), claim your free points, and forecast YES or NO on an open claim. There's no gas to pay: our relayer submits it.
• Watch the whole cycle: on Try it now, open a demo claim with made-up data, forecast on it, and watch the Lab retrace it live a few minutes later.
• Check a paper's statistics, no wallet: Check a paper → Try an example.
• From a terminal, no wallet or keys: with Node ≥ 22.22.2 and pnpm 12, run pnpm install, then pnpm judge-demo. It makes a throwaway testnet account, claims points, forecasts, re-checks a finished receipt, and tries to cheat, printing PASS or BLOCKED for each.

What the Lab found

Published results, rerun from public code and data. The forecast price never changes a verdict: plain code compares the Lab's number with the paper's, within a margin fixed before anyone forecast.

여기까지가 README 앞부분입니다. 나머지는 원문에서 읽을 수 있습니다. README 전문 보기 ↗