OpenAI claims a general-purpose reasoning model found a counterexample to Erdos's unit-distance bound [D]

2026-05-20 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Advanced, medium

Summary

OpenAI announced that one of its general-purpose reasoning models discovered a construction that disproves the conjectured n^{1+O(1/log log n)} upper bound in Erdős's planar unit-distance problem. This mathematical claim suggests the existence of finite planar point sets with more than n^{1+δ} unit distances for some fixed δ > 0 and infinitely many n, challenging the expected near-linear upper bound. The company states the solution was generated by a general-purpose reasoning model, then validated by an AI grading pipeline and human mathematicians. However, OpenAI's disclosure lacks crucial experimental details, including the model name, sampling setup, number of attempts, compute budget, hidden system prompt, or full grading pipeline, leading to community debate on the AI's autonomy and the reproducibility of the result.

Key takeaway

For research scientists evaluating AI's role in mathematical discovery, you should view OpenAI's claim as a significant demonstration of AI's capability to generate complex mathematical constructions. However, demand full methodological transparency, including model specifics and attempt statistics, before treating this as a reproducible milestone. Your ability to assess the true autonomy and generalizability of such AI systems depends on complete experimental details.

Key insights

A general-purpose AI model found a counterexample to a long-standing mathematical conjecture, highlighting AI's potential in complex problem-solving.

Principles

• AI can exploit known results from unrelated fields for novel combinations.
• Disproving conjectures may be an area where AI excels.
• Full methodological disclosure is crucial for AI-for-math reproducibility.

Method

A general-purpose reasoning model generated a solution, which was then checked by an AI grading pipeline and refined by mathematicians. The initial prompt was also AI-generated.

In practice

• Use AI to explore counterexamples in mathematical conjectures.
• Combine AI-generated prompts with reasoning models.
• Employ AI grading pipelines for solution verification.

Topics

• General-Purpose AI
• Mathematical Discovery
• Erdős Unit-Distance Problem
• AI Reproducibility
• Discrete Geometry
• Model Transparency

Best for: AI Scientist, Research Scientist, Tech Journalist

Related on AIssential

• OpenAI Says Its AI Just Disproved a Famous 80-Year-Old Math Problem – And This Time, Mathematicians Agree — A general-purpose AI autonomously disproved an 80-year-old math conjecture by making a cross-domain mathematical leap.
• OpenAI reasoning model disproves 80-year-old Erdős geometry conjecture — A general-purpose AI model has autonomously disproved an 80-year-old mathematical conjecture, demonstrating advanced reasoning capabilities.
• OpenAI just SOLVED MATH.... — AI's breakthrough stemmed from its ability to connect disparate mathematical fields, revealing novel solutions.
• An AI solution to an 80-year-old problem has shocked mathematicians — AI can autonomously solve long-standing mathematical conjectures by exploring vast solution spaces and combining existing knowledge.
• OpenAI claims it solved an 80-year-old math problem — for real this time — OpenAI's new model disproved an 80-year-old math conjecture, demonstrating advanced autonomous reasoning in AI.
• AI cracks 80-year-old mathematics challenge — researchers are astonished — AI can autonomously solve complex, long-standing mathematical conjectures.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.