OpenAI System Resolves Longstanding Unit Distance Problem In Geometry

A long-standing problem in discrete geometry involving how many pairs of points can be exactly one unit apart has been addressed using an artificial intelligence system developed by OpenAI, according to company research published on its official site

The problem, known as the planar unit distance problem, asks for the maximum number of pairs at distance one among n points in a plane. It was first introduced in 1946 by Hungarian mathematician Paul Erdős and has remained one of the most widely studied questions in combinatorial geometry.

For decades, mathematicians worked on the assumption that grid-based arrangements of points were close to optimal for producing unit distances. These configurations were believed to provide only marginal improvements over linear growth patterns.

OpenAI stated that an internal reasoning model has now produced a proof that challenges that assumption. The model generated an infinite family of geometric constructions that achieve a polynomial improvement over previously known bounds. The company also said the proof has been reviewed and checked by external mathematicians, along with a companion paper explaining the method and context.

The unit distance problem has been a central topic in combinatorial geometry for decades. It examines how point arrangements behave under fixed distance constraints and has been part of ongoing mathematical research since Erdős first proposed it in the mid-20th century. Additional background on earlier approaches, including grid-based constructions and known bounds, is documented in Wikipedia.

According to the OpenAI report, the new proof shows that for infinitely many values of n, there exist configurations of points that exceed previously assumed limits on unit-distance pairs. The result contradicts a widely held belief that growth in such configurations could not significantly surpass near-linear rates.

The proof was independently verified by external mathematicians, who also contributed a companion analysis discussing its structure and mathematical implications. Fields Medalist Tim Gowers described the result as a notable milestone in AI-assisted mathematics, while number theorist Arul Shankar said the model demonstrated the ability to generate original mathematical ideas and carry them through complete proofs.

The construction used in the proof draws on algebraic number theory, a branch of mathematics that studies number systems extending the integers. These systems include algebraic number fields, which have been used in earlier partial approaches to geometric problems.

Earlier approaches to the unit distance problem often used structures such as Gaussian integers to build dense point configurations. The new result extends these ideas using more complex algebraic constructions, allowing for denser arrangements than previously established methods, according to Wikipedia.

The OpenAI report also states that the proof emerged from a general-purpose reasoning model rather than a system specifically designed for geometry or formal theorem-proving. The company said the model was tested across a set of problems associated with Erdős and produced the solution as part of that evaluation.

External mathematicians involved in reviewing the result stated that the argument connects algebraic number theory with discrete geometry in a way not previously used in this context. The companion paper provides additional details on how the construction was verified and analyzed.

The development adds to growing interest in the use of AI systems in mathematical research. OpenAI said the result marks a step in demonstrating that advanced AI systems can contribute to resolving long-standing open problems in mathematics, particularly in areas involving complex logical reasoning and multi-step proofs.