OpenAI Model Disproves 80-Year Unit Distance Conjecture With n^(1+0.014) Construction
Updated
Updated · OpenAI · May 20
OpenAI Model Disproves 80-Year Unit Distance Conjecture With n^(1+0.014) Construction
5 articles · Updated · OpenAI · May 20
An OpenAI reasoning model produced a proof overturning Erdős’s 1946 unit distance conjecture, showing that for infinitely many n, point sets can realize at least n^(1+δ) unit-distance pairs.
A forthcoming refinement by Princeton’s Will Sawin makes the gain explicit with δ=0.014, beating the long-prevailing view that square-grid-style constructions were essentially optimal at only n^(1+o(1)).
External mathematicians checked the proof and wrote a companion paper, while OpenAI said the model was general-purpose rather than trained specifically for mathematics or this problem.
The argument imports algebraic number theory tools—including class field towers and Golod–Shafarevich theory—into an elementary geometric question, a link mathematicians called unexpected and potentially useful for related problems.
OpenAI and outside researchers framed the result as the first autonomous AI solution to a prominent open problem in an active math subfield, pointing to broader research uses if such reasoning survives expert scrutiny.
What does AI's alien-like mathematical intuition reveal about the unseen limits of human creativity?
With math conquered, which scientific field will be the next to be completely transformed by AI?
When AI's 'genius' solves a famous problem, how do we separate breakthrough from hallucination?
OpenAI Model Breaks 80-Year-Old Erdős Unit Distance Conjecture, Redefining AI’s Role in Mathematics
Overview
On May 14, 2026, an internal OpenAI model achieved a major breakthrough by autonomously disproving the 80-year-old Erdős unit distance conjecture, a problem first proposed by Paul Erdős in 1946. This milestone highlights artificial intelligence’s growing ability to solve creative and complex problems, generating new insights in pure mathematics. For decades, mathematicians struggled with the conjecture, establishing various bounds but never settling the question. The AI’s success not only marks a significant advance for AI research but also demonstrates its potential to contribute original ideas and reshape the future of mathematical discovery.