Proposal: Community Notes-style resolution for vague predictions on X

Raemon, a LessWrong moderator, outlined a proposal to integrate prediction tracking directly into Twitter/X, aiming to raise 'median sanity' on a global scale. The idea: any tweet that makes a prediction is automatically converted into a structured prediction object. An AI suggests a few plausible operationalizations and likely edge cases, making vague claims testable without requiring users to write precise criteria themselves.

Instead of formal resolution, the system would use a Community Notes-style voting mechanism where users decide if a prediction resolved true, with algorithms designed to remain robust to partisan disagreements. Popular predictions would automatically receive AI-generated evidence collection and citations, giving readers a percentage score alongside concrete facts. Users earn badges and leaderboard rankings for making accurate predictions. Raemon notes prediction markets (like Manifold or Metaculus) incentivize arbitrage rather than useful predictions, but believes those platforms could prototype the UI. He also highlights X API costs as a barrier and wonders if Musk, who still trusts Community Notes, would champion this extension.