In 1946, Paul Erdős posed a deceptively simple question that would haunt mathematicians for nearly eight decades: if you place dots on a piece of paper, how many pairs can be exactly the same distance apart? The legendary Hungarian mathematician proposed an answer—that the maximum grows only slightly faster than the number of dots themselves—but no one could prove it. Until now, when an artificial intelligence system working inside OpenAI's labs accomplished what human experts could not, suggesting that one of the 20th century's greatest mathematicians may have gotten it wrong.
The breakthrough hinges on what's called the planar unit distance problem, a corner of geometry that has resisted proof through traditional mathematical approaches. Erdős's conjecture felt intuitively correct to many researchers, and for good reason—his track record was formidable. Yet without a rigorous proof, the question sat unresolved, a small but nagging gap in mathematical knowledge. Understanding how many equal distances can exist within a point set has applications beyond pure geometry, touching optimization problems and algorithmic design. This is why the breakthrough matters: it's not merely academic curiosity, but a test case for whether AI can genuinely contribute to unsolved mathematical problems.
What makes the solution remarkable is its origin. OpenAI researchers simply fed the Erdős conjecture into a large language model and asked whether the mathematician was wrong. The AI didn't attack the problem directly with standard geometric approaches. Instead, it drew on multiple branches of mathematics—employing complex number systems and higher-dimensional geometric thinking—to explore the question from unexpected angles. The result was hundreds of pages of calculations that led to a startling conclusion: there is no universal speed limit as Erdős had proposed. Instead, you can arrange dots in ways that produce vastly more equal distances than anyone previously thought possible.
The findings have been rigorously verified. Expert mathematicians checked the proof and confirmed the core discovery, formally disproving the unit distance conjecture with mathematical precision. The researchers themselves acknowledge the unconventional path: they were not directly involved in the AI's initial calculations but stepped in afterward to review the software's logic and help rewrite the findings for clarity and rigor. This collaboration between machine computation and human mathematical judgment proved essential.
What makes this moment stand out is the transparency of the process. Rather than a black box producing an answer, the AI generated a verifiable proof that experts could examine line by line. The breakthrough has been formally documented both in a mathematical proof and in a companion paper on the arXiv preprint server, making it available for scrutiny across the global mathematics community. This openness is crucial—it allows the breakthrough to be tested, challenged, and ultimately trusted.
For those who work in mathematics, the implications reach beyond Erdős's specific puzzle. This represents one of the clearest examples to date of AI genuinely helping to crack a longstanding mathematical problem that had stumped leading experts. It suggests that large language models can do more than process language: they can reason through abstract mathematical structures, explore novel solution strategies, and contribute meaningfully to human knowledge. The question now is not whether AI can solve hard math problems, but how many more have been waiting all along for the right tool to ask them the right question.