Mathematicians spent nearly eight decades attempting to resolve the inquiry. The ‘planar unit distance problem’ is a well-known Maths enigma that was first posed by the iconic mathematician Paul Erdős in 1946, later becoming one of the most recognized unsolved puzzles in geometry.
“If you scatter numerous dots on a flat plane, what is the maximum number of dot pairs that can be precisely one unit apart from each other?” A ‘unit’ could represent one centimeter, one inch, or any fixed length. This challenge revolves around positioning dots on a flat surface in the most efficient manner.
Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946.
For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids.
An OpenAI model has now disproved that… pic.twitter.com/j2g3Ze0zEG
— OpenAI (@OpenAI) May 20, 2026
For many years, experts posited that the optimal arrangement resembled a square grid—akin to dots systematically aligned in rows and columns, much like graph paper. This approach yielded numerous pairs of dots that were exactly one unit apart, leading mathematicians to believe it was likely the best possible configuration.
Now,
OpenAI asserts that its AI reasoning model has discovered a novel arrangement of points that surpasses the traditional square-grid concept.
OpenAI indicates this new pattern represents a ‘polynomial improvement,’ signifying that the enhancement grows significantly as the number of dots increases.
Tim Gowers, a recipient of the Fields Medal, described it as “a milestone in AI Mathematics.” Prominent mathematician Arul Shankar noted that this achievement demonstrates AI systems’ capacity to generate “original ingenious ideas” rather than merely assisting humans with calculations.
The proof also garnered admiration from researchers, as the AI employed sophisticated concepts from algebraic number theory to tackle what initially seemed like a straightforward geometry problem.
Typically, AI systems working on Mathematical problems are specifically trained for those tasks and guided step-by-step by humans. Researchers were exploring whether advanced AI models could contribute to challenging scientific inquiries, and during testing, the model produced a proof for this historically unsolved Maths problem.
“The frontiers of knowledge are very spiky, and no doubt the coming months and years will witness similar breakthroughs in various fields of mathematics, where long-standing open questions are resolved by an AI uncovering unforeseen connections and pushing existing technical limits,” remarked mathematician Thomas Bloom.
