OpenAI Model Disproves Erdős Unit-Distance Conjecture With New Proof
Google DeepMind/Pexels
Source Analysis
What Happened
Why It Matters
What Happened
Why It Matters
Where Sources Disagree
- arrows_outputOriginality of AI's Proof: OpenAI and some mathematicians assert the AI generated genuinely original mathematical ideas and a novel, significant proof. However, other mathematicians and researchers contend the solution was not fundamentally new, executing existing ideas.
- arrows_outputOpenAI Proof Validation: Some outlets emphasize OpenAI's announcement faced heavier scrutiny due to past misrepresentations. In contrast, other outlets highlight the proof was simply verified by external mathematicians, a standard practice.
Timeline
May 21, 2026
Mathematicians Generalize And Improve: In the days after the release, researchers such as Will Sawin and others refined and extended the AI-derived construction—expressing the improvement with a fixed exponent and producing slight further improvements using the newly revealed algebraic-number-theory approach. Those follow-ups showed the technique could be understood, internalized, and built upon by human mathematicians.
May 21, 2026
External Verification And Commentary: Following the announcement, OpenAI solicited and published companion remarks and independent reviews from several mathematicians and produced a companion paper; external experts affirmed the argument's novelty and rigor while emphasizing peer review as the credibility standard. Mathematicians including Noga Alon, Melanie Wood, Thomas Bloom and Tim Gowers provided public commentary and participated in verification and companion analysis.
May 20, 2026
OpenAI Announces Disproof: On May 20, 2026, OpenAI announced that an internal general-purpose reasoning model produced an original proof that disproves Erdős's unit-distance conjecture by identifying an infinite family of point constructions that yield substantially more unit-distance pairs. The company published a long chain-of-thought account and said the proof came from a general-purpose model rather than a specialized theorem prover.
Summary by Ground AI
Sources
See All 45A varied selection of sources chosen by Ground to reflect the diversity of this story’s coverage.
Most Written About
The stories getting the most coverage from the last 24 hours
Source Analysis
Timeline
May 21, 2026
Mathematicians Generalize And Improve: In the days after the release, researchers such as Will Sawin and others refined and extended the AI-derived construction—expressing the improvement with a fixed exponent and producing slight further improvements using the newly revealed algebraic-number-theory approach. Those follow-ups showed the technique could be understood, internalized, and built upon by human mathematicians.
May 21, 2026
External Verification And Commentary: Following the announcement, OpenAI solicited and published companion remarks and independent reviews from several mathematicians and produced a companion paper; external experts affirmed the argument's novelty and rigor while emphasizing peer review as the credibility standard. Mathematicians including Noga Alon, Melanie Wood, Thomas Bloom and Tim Gowers provided public commentary and participated in verification and companion analysis.
May 20, 2026
OpenAI Announces Disproof: On May 20, 2026, OpenAI announced that an internal general-purpose reasoning model produced an original proof that disproves Erdős's unit-distance conjecture by identifying an infinite family of point constructions that yield substantially more unit-distance pairs. The company published a long chain-of-thought account and said the proof came from a general-purpose model rather than a specialized theorem prover.
