OpenAI's reasoning model autonomously disproved a conjecture posed by legendary mathematician Paul Erdős in 1946, and top mathematicians say the proof is publishable on its own merits.
For nearly eight decades, some of the sharpest minds in mathematics stared at a deceptively simple question about dots on a flat surface and couldn’t crack it. An AI just did. OpenAI announced that one of its internal reasoning models autonomously disproved the planar unit distance conjecture, a problem first posed by Hungarian mathematician Paul Erdős in 1946. The conjecture asked a straightforward-sounding question: what’s the maximum number of pairs of points you can place exactly one unit apart on a two-dimensional plane? Turns out the answer everyone assumed was right, wasn’t.
What the AI actually proved #
For decades, mathematicians believed the optimal arrangement of points involved square-grid configurations. The AI produced a new family of geometric constructions that outperform the square-grid approach, effectively disproving the conjecture. The proof draws on infinite class field towers and the Golod-Shafarevich theorem. Leading mathematicians have independently verified the proof. Fields Medalist Timothy Gowers described the approach as “clever” and “elegant.” Other reviewers stated the proof could be published in a top-tier mathematical journal on its own merits.
Why this is different from previous AI math achievements #
The AI pulled from algebraic number theory (class field towers) to solve a problem in combinatorial geometry, a cross-disciplinary leap that most human mathematicians wouldn’t instinctively attempt.
Paul Erdős posed over 1,500 problems across his career and offered cash prizes for their solutions. The planar unit distance conjecture was one of his more stubborn legacies.
The announcement was made public between May 20 and May 21, 2026.
What this means for investors #
OpenAI’s reasoning models are demonstrating capabilities that were considered science fiction even two years ago. Autonomously producing publishable-quality mathematical proofs, the kind that earn Fields Medalists’ praise, represents a step change in what AI systems can do with abstract reasoning and logical creativity.
The same reasoning capabilities that crack open geometry problems can be applied to cryptography, protocol design, smart contract verification, and risk modeling. If AI can navigate the abstract landscape of class field towers and the Golod-Shafarevich theorem, formal verification of complex DeFi protocols starts looking a lot more tractable.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our