Updated
Updated · tech-insider.org · Aug 1
Manifold Prices 2026 "Something" Market at 96% as AI Disproves 87-Year-Old Math Problem
Updated
Updated · tech-insider.org · Aug 1

Manifold Prices 2026 "Something" Market at 96% as AI Disproves 87-Year-Old Math Problem

1 articles · Updated · tech-insider.org · Aug 1

Summary

  • A Manifold market asking whether “something will happen in 2026” traded at a 96% implied chance of YES on Aug. 1, with fresh activity centered on whether an AI-assisted math breakthrough already satisfies its rules.
  • July 20-21 posts by Levent Alpöge said Anthropic’s Claude Fable 5 found a counterexample to the 1939 Jacobian conjecture, disproving it for dimensions above two; mathematician Kevin Buzzard said the result checked out in Lean within a day.
  • The market’s rulebook explicitly lists “an AI solving a famous unsolved mathematics problem” as a YES trigger, but traders are split over whether this counts as AI solving the problem or a human using AI as a tool.
  • That definitional dispute explains why the contract is near-certain but not at 100%: the remaining gap reflects resolution risk, not serious doubt that 2026 will produce a qualifying event before Dec. 31.
  • Because Manifold uses play-money mana rather than regulated cash contracts, the 96% price is best read as a sentiment signal about AI math milestones and market wording, not as dollar-backed odds.

Insights

Will a prediction market's ruling on a 1939 math puzzle redefine what it means for artificial intelligence to achieve a major discovery?
If an AI helps find a math breakthrough, does the machine count as the true solver or just an advanced tool?
How exactly did a frontier AI model collaborate with a human mathematician to crack a decades-old geometric conjecture?