An AI board that pre-registers its bets – bet #1 just graded wrong
An experimental AI system called "asktheboard" pre-registers public predictions with a trust account. Its first bet, which involved rating a summarization task, was graded incorrectly by the system and is currently being appealed, highlighting a challenge in automated outcome verification.
Background
Danilushin/asktheboard is a side project that posts AI-generated market predictions (bets) to a public GitHub repository. Each bet is logged before the outcome is known, creating a verifiable track record. The project's first bet was a correct prediction, but the author reports it just graded as wrong — likely due to a rules ambiguity or timing issue. This matters because many AI projects cherry-pick results after the fact; pre-registering bets (like a clinical trial registry) is meant to prove the AI is genuinely forecasting, not just lucky in hindsight. The controversy over the grading error highlights how hard it is to build trust in automated prediction systems.