The date of May 20, 2026, is regarded as a milestone in the fields of mathematics and artificial intelligence. OpenAI's internal reasoning model successfully refuted the "Unit Distance Problem," proposed by the mathematical master Paul Erdős in 1946, which had been a core open problem in discrete geometry for nearly 80 years.

image.png

Key Breakthrough: From "Retrieval" to "Originality"

Differing from previous AI mathematical capability claims that sparked controversy, this achievement has received widespread endorsement from the academic community.

  • Non-retrieval Solution: This proof was not compiled by the model from historical literature, but rather the model independently discovered a new construction method based on number theory (particularly algebraic number field construction). This approach demonstrated the existence of point sets in a plane with unit distance pairs exceeding the previously believed optimal upper limit, directly refuting Erdős's conjecture.

  • Expert-Level Validation: The discovery was then reviewed in detail by a team of renowned mathematicians, including Thomas Bloom, Noga Alon, and Tim Gowers. Mathematician Tim Gowers explicitly stated that if this paper were written by a human and submitted to the "Annals of Mathematics," he would have no hesitation in recommending its acceptance.

Technical Significance: Evolution of AI Reasoning Ability

  • Long-Chain Logical Ability: Solving this problem required the model to demonstrate the ability to handle long-chain reasoning. The mathematical proof process involves hundreds of rigorous logical deductions, not only requiring high logical coherence but also creative exploration in the face of open problems, rather than simply repeating patterns in training data.

  • Not Just a Tool, but a Partner: This event marks AI's evolution from a "auxiliary computing tool" into an "original research partner." It can perform computations, build theories, and propose mathematical paths that humans had never imagined before.

Application Impact: Beyond Mathematics Itself

Research findings on the unit distance problem often have broad implications. Understanding efficient arrangement patterns of point sets in space has profound practical value for the following fields:

  • Materials Science: Optimizing the design of crystal structures;

  • Engineering and Communication: Network topology optimization and wireless communication system design;

  • Biomedical: Molecular design, protein folding, and simulation of biomolecular structures.

Historical Echoes and Warnings

This achievement has been likened to the moment in 1976 when the computer-assisted proof of the "Four Color Theorem" was announced. Although OpenAI had previously faced controversies over "exaggerated promotion" in the field of mathematics, the rigor and originality of this proof have been publicly released on arXiv (arXiv:2605.20579v1) and have undergone peer review by the global mathematics community.