OpenAI's AI Resolves an 80-Year-Old Unsolved Math Problem

In mid-May 2026, OpenAI made a stunning announcement: its internal AI model had disproven the famous unsolved problem in discrete geometry, the “Erdős Unit Distance Conjecture.” This problem, posed by the brilliant mathematician Paul Erdős in 1946, had confounded mathematicians worldwide for 80 years. This achievement, heralding a new era of AI-driven mathematical proofs, has sent ripples through the academic community.

A Landmark Recognized by Leading Mathematicians

To validate its findings, OpenAI provided early access to several prominent mathematicians, including recipients of the Fields Medal, often regarded as the “Nobel Prize of Mathematics.”

Fields Medalist Tim Gowers commented, “Resolving the unit distance problem is undoubtedly a milestone in AI mathematics.” Likewise, Professor Daniel Litt of the University of Toronto remarked, “This is the first instance where I’ve found an AI-derived result to be inherently fascinating, not just as a technological breakthrough but as a result in its own right.”

Indeed, this may well be the first time an AI system has discovered a proof for a major unsolved mathematical conjecture. In this sense, OpenAI’s achievement is undeniably remarkable.

The Trajectory of AI in Mathematics

While groundbreaking, this achievement is more of a steady progression in AI’s evolution in mathematics than a sudden leap forward.

Just three years ago, large language models (LLMs) struggled with basic arithmetic problems. It was only last year that such systems began achieving high scores in high school-level math competitions.

At the “Joint Mathematics Meetings,” the world’s largest annual mathematics conference held in January 2026, the author learned that while AI systems were beginning to contribute to mathematical research, their applications remained confined to limited scenarios. Significant human interpretation and adjustment were still required to turn AI outputs into publishable theorems.

OpenAI’s Achievement as a New Milestone

OpenAI’s recent success represents the next critical step in this evolutionary process. The AI model skillfully applied existing ideas from multiple subfields of mathematics to construct a complete proof. This achievement vividly demonstrates the improving ability of AI to emulate and combine mathematicians’ knowledge and creativity.

However, it should be noted that the model did not pioneer entirely new mathematical methods. Subsequent reports revealed that human mathematicians further organized and expanded upon the proof. This highlights an emerging collaborative model for future mathematical research involving both humans and AI.

What Is the Erdős Unit Distance Problem?

To understand the core of this breakthrough, it’s essential to grasp the essence of the Erdős Unit Distance Conjecture.

Paul Erdős, one of the most prolific mathematicians in history with over 1,500 published papers, was renowned for posing seemingly simple yet deeply complex mathematical problems.

The unit distance problem, introduced by Erdős in 1946, poses the following question:

Imagine a set of points on a two-dimensional plane. Measure the distances between all pairs of points in the set. Can the number of point pairs that are exactly “one unit” apart be maximized by adjusting the arrangement of the points? And what is the maximum ratio of these unit-distance pairs to the total number of points?

To illustrate, suppose there are five points on a plane, which form 10 total pairs. In one specific arrangement, three pairs might be exactly one unit apart. Is it possible to rearrange the points to create even more pairs at a one-unit distance? While straightforward to state, solving this problem requires intricate considerations of discrete geometry.

The AI’s Approach to the Proof

OpenAI’s AI model constructed a disproof of this conjecture, effectively overturning long-held assumptions.

As the article’s author points out, the strength of this AI model lies in its vast knowledge of past mathematical research and its capacity to persistently attempt tedious, repetitive proof strategies that humans might avoid. Rather than relying on human-like intuition or creativity, the AI achieved its breakthrough through exhaustive exploration and the combination of existing methods.

This outcome suggests that AI is beginning to take on a role in mathematics that leverages humanity’s accumulated knowledge and computational prowess.

The Long-Term Future of Human-AI Collaboration

This achievement points to a future where human mathematicians and AI models coexist, each leveraging their unique strengths.

AI has access to a broader range of prior research knowledge than any living human and possesses the patience to diligently explore even the most tedious proof strategies. On the other hand, humans excel at deeply contemplating specific problems and posing intriguing questions that AI might not conceive on its own.

This collaborative relationship holds the potential to dramatically enhance productivity in mathematical research. AI can broaden the exploration space, identify promising candidates, and leave it to humans to discern the essence and deepen the understanding of these findings. Such a future is becoming increasingly plausible.

Will AI Replace Human Mathematicians?

However, the longevity of this collaborative model is uncertain. AI systems continue to evolve rapidly in the field of mathematics. The article’s author notes, “In 10 years, it is unclear what role human mathematicians will play.”

This suggests that AI’s capabilities are advancing beyond mere computational speed, extending to devising proof strategies and gaining fundamental insights into problems.

That said, current AI has not yet reached the stage where it can generate genuinely new mathematical concepts. While it excels at applying and combining existing knowledge, it has yet to replicate the human capacity for mathematical intuition and aesthetic judgment. This will likely be an important area of focus for future AI research.

Democratizing and Accelerating

Mathematical Research

Automating mathematical proofs with AI may not only push the frontiers of research but also transform the nature of mathematical inquiry itself.

For instance, researchers without advanced mathematical expertise might use AI tools to validate complex proofs or explore new hypotheses. Meanwhile, by delegating tedious calculations to AI, human mathematicians could concentrate on more creative aspects of their work.

The resolution of the Erdős Unit Distance Conjecture might represent the first significant step toward this visionary future.

Conclusion:

The Dawn of a New Era in AI Mathematics

OpenAI’s disproof of the Erdős Unit Distance Conjecture is a landmark moment, signaling that AI has begun to rival—and even surpass—humans in advanced mathematical reasoning.

For 80 years, this problem has stymied mathematicians worldwide. Its resolution by AI is not merely a victory of computational power but a testament to the integration of vast knowledge and logical reasoning.

This achievement indicates that AI mathematics has entered the “practical application” phase. Ahead lies a new era where humans and AI collaborate, leveraging their respective strengths to explore uncharted mathematical territories.

However, the pace of AI’s evolution remains unpredictable, and the role of human mathematicians may undergo significant changes. What is certain is that mathematics, one of humanity’s oldest disciplines, has gained its most powerful partner yet in AI, propelling its frontiers forward at an unprecedented pace.