OpenAI model solves 80-year-old planar unit distance problem

For 80 years, one of the most stubborn problems in combinatorial geometry sat on the shelf, occasionally dusted off by ambitious mathematicians, never quite cracked. Now an AI did it.

An internal OpenAI general reasoning model has produced a proof that resolves the planar unit distance problem, a conjecture first posed by legendary Hungarian mathematician Paul Erdős in 1946. The proof, spanning roughly 125 pages, establishes an infinite family of planar configurations with more unit-distance pairs than the traditionally assumed optimal arrangements. In plain terms: the AI found geometric patterns that break a limit mathematicians believed held for eight decades.

What the proof actually says

The planar unit distance problem asks: given n points in a plane, what is the maximum number of pairs that can be exactly one unit apart? Erdős conjectured an upper bound on this count, and for decades, the best-known configurations were grid-like structures that seemed to confirm his intuition.

OpenAI’s model took a different route entirely. Rather than iterating on known grid arrangements, it approached the problem through algebraic number theory, connecting it to advanced mathematical structures called infinite class field towers. The result is an infinite family of configurations that surpass the traditionally accepted optimal ones, refuting Erdős’s conjectured upper bound outright. The improvement has been quantified with an exponent of approximately 0.014.

Who verified it, and why that matters

Tim Gowers, a Fields Medalist, reviewed the work. So did Will Sawin, a mathematician at Princeton. Both validated the proof’s correctness. Sawin specifically quantified the improvement at the roughly 0.014 exponent figure.

The announcement came around May 20, 2026, and it has immediately reshaped the conversation about what AI reasoning systems can do in pure research contexts.

What this means beyond mathematics

The techniques involved, particularly the algebraic number theory and the construction of novel mathematical objects, have direct relevance to formal verification and zero-knowledge proof systems.

Formal verification is the process of mathematically proving that code does what it’s supposed to do. If AI reasoning models can generate and validate proofs at the level demonstrated here, the cost and timeline for formally verifying complex protocols could drop dramatically.

Zero-knowledge proofs, the cryptographic technique underpinning privacy-focused blockchains and scaling solutions like zk-rollups, are built on deep algebraic foundations. The kind of algebraic number theory OpenAI’s model employed to crack this problem lives in the same mathematical neighborhood.

No specific crypto tokens are tied to this result, and anyone claiming otherwise is getting ahead of the facts.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.