Will any of OpenAI’s massive release of mathematical results be shown to be flawed by the end of November?
AIOne-Off
Manifold MarketsClosedNo KYCVerified resolution dataFairly calibratedThis market resolved: Yes (99%)
Resolved: Oct 8, 2026, 11:37 AM
Provider-verified resolution
CoinRithm has provider-backed winner and settlement-time evidence for this market.
Winning Outcome
Yes
At 7 days out
—
At 24 hours out
37%
SurpriseData quality warningStale data
Data as of Oct 8, 2026, 9:51 PM UTC · policy pm-quality-4
Probability Timeline
Yes99%
No1%
Won
Lost
Related Markets
Related questions for context, not a comparison of equivalent contracts.
Which of OpenAI's claimed math results will be shown to be flawed before 2027?
Manifold MarketsWill any of Levent Alpöge’s mathematical results with Claude be proven flawed by the end of the year?
Manifold MarketsOpenAI's proof of zero free region of Riemann zeta for Re(s)>7/8 shown to be flawed before 2027?
Manifold MarketsBest AI at the end of 2026?
KalshiWhich company has the best AI model end of October?
PolymarketWhich company has best AI model end of 2026?
PolymarketRules
[tweet]YES if, by November 30, there is a consensus in credible reporting or within the academic community that at least one of the core arguments, theorems, or proofs in OpenAI's GitHub repo is fatally flawed or incorrect.
- Otherwise, resolves NO.
- "Flawed" means a core mathematical or theoretical error is identified that invalidates the main result of the paper, and the error is either unfixable or difficult to patch.
- Nitpicks—such as minor typos, formatting errors, minor translation gaps in the Lean formalization that do not undermine the mathematical proof's validity, or easily patchable edge cases—do not count toward a YES resolution.
- The flaw has to be in the repo as it is released today - later hotfixes don’t remove an eligible flaw, but later additions don’t count either.