LOADING...
OpenAI says its new AI model solved decades-old geometry conjecture
The mathematical question was first posed in 1946

OpenAI says its new AI model solved decades-old geometry conjecture

May 21, 2026
12:38 pm

What's the story

OpenAI claims that its new general-purpose reasoning model has successfully disproved a famous unsolved conjecture in geometry, the planar unit distance problem. The mathematical question was first posed by legendary mathematician Paul Erdos in 1946. It asks how many pairs of points on a flat surface can be exactly one unit apart from each other. For nearly eight decades, mathematicians assumed the best arrangements looked like square grids.

Breakthrough achievement

AI discovered a more efficient way to arrange points

OpenAI claims its new general-purpose reasoning model has disproved the long-standing assumption by finding a new family of mathematical constructions that perform better than the traditional grid-like approach. This means the AI discovered a more efficient way to arrange points than humans had thought possible for decades. The result is not just significant because of the solution itself but also due to how it was achieved.

Verification process

Proof generated by the model was later verified by mathematicians

The proof generated by OpenAI's general-purpose reasoning model was later verified by external mathematicians. This is the first time an AI system has autonomously solved a prominent open problem at the heart of an active field of mathematics. For many researchers, this milestone is as important as the mathematical discovery itself. It also shows how quickly AI reasoning capabilities are evolving beyond just generating text, images, or computer code.

Advertisement

Future implications

AI systems are beginning to independently explore ideas, uncover solutions

OpenAI argues that if AI systems can sustain difficult reasoning and produce work that survives expert scrutiny, they could eventually help scientists tackle complicated problems in biology, physics, engineering, materials science, and medicine. The development also highlights a growing shift in the relationship between humans and artificial intelligence. Instead of merely assisting researchers with calculations or data analysis, AI systems are beginning to independently explore ideas and uncover solutions experts themselves may not have considered.

Advertisement