OpenAI AI Disproves Erdős's 1946 Conjecture, Cracking an 80-Year-Old Geometry Problem
Updated
Updated · The Conversation · May 26
OpenAI AI Disproves Erdős's 1946 Conjecture, Cracking an 80-Year-Old Geometry Problem
7 articles · Updated · The Conversation · May 26
OpenAI said one of its general-purpose AI models found a counterexample to Paul Erdős’s 1946 planar unit distance conjecture, overturning a belief that had guided work on the problem for 80 years.
The proof shows that, for infinitely many values of n, point arrangements drawn from algebraic number theory produce more unit-distance pairs than the square-grid style constructions long thought near-optimal.
Daniel Litt called it the first autonomous AI result he found intrinsically interesting, while Fields medallist Timothy Gowers said he would back publication in Annals of Mathematics “without any hesitation.”
Will Sawin quickly extended the same line of reasoning to a stronger result, though the improvement reportedly appears only around 10^2000000 points, underscoring both the breakthrough and its extreme scale.
The result is being treated as a milestone for AI-assisted mathematics because OpenAI says the model worked with minimal human intervention, even as mathematicians still question whether AI can supply truly original conceptual leaps.
An AI just disproved a famous math conjecture. Which long-standing scientific belief will be the next to fall?
AI is authoring breakthrough proofs, but with a 25% error rate. Are we entering a golden age or an era of 'AI slop'?
As AI solves problems that stumped humans for 80 years, what is the new role for human creativity in science?
OpenAI’s AI Model Solves 80-Year-Old Erdős Unit Distance Problem, Redefining Mathematical Research (May 2026)
Overview
In May 2026, OpenAI's AI model made history by autonomously disproving Paul Erdős's unit distance conjecture, a problem that had challenged mathematicians for nearly 80 years. The AI achieved this by generating point arrangements that broke the long-accepted theoretical limits, using advanced mathematical tools like algebraic number theory and high-dimensional lattices. Human experts validated the result, and the breakthrough sparked excitement in the scientific community, highlighting AI's creative potential and the growing importance of human-AI collaboration. This achievement is reshaping research workflows, raising new standards for verifying AI-generated mathematics, and signaling a new era where AI accelerates scientific discovery.