Summary
OpenAI has achieved a genuine mathematical breakthrough: its general-purpose reasoning model has disproved a conjecture posed by legendary Hungarian mathematician Paul Erdős in 1946 known as the planar unit distance problem. The conjecture asked how many pairs of points at exactly the same distance apart can exist given a set of dots on a plane — Erdős believed the number would rise only slightly faster than the number of dots. OpenAI’s model proved him wrong.
The AI discovered an entirely new family of geometric constructions that outperform what mathematicians had believed were optimal square-grid-like arrangements for nearly eight decades. Crucially, this result has been independently validated by mathematicians, including Thomas Bloom, who maintains the official Erdős Problems archive and had previously criticized OpenAI’s earlier Erdős claims. Bloom co-authored a companion paper confirming the discovery.
Source
Commentary
This is significant because it addresses the most damning criticism of AI in mathematics: that models merely regurgitate existing literature rather than generate novel insights. OpenAI’s previous Erdős claim last year was exactly that — recycled results dressed up as breakthroughs. This time, independent mathematicians have confirmed the result is genuinely new.
Bloom’s observation that the AI succeeded by “persevering down paths that a human may have dismissed as not worth their time” is perhaps the most interesting takeaway. It suggests AI’s mathematical value may not be in brilliance but in sheer exhaustive patience — exploring solution spaces that humans would rationally skip. That’s a legitimate and powerful capability, even if it’s less romantic than the “AI genius” narrative.