Autonomous disproofs of the Erdős unit distance conjecture with GPT-5.5 Pro

OpenAI's disproof of the Erdős unit distance conjecture marks a milestone in AI for mathematical research. However, the result relied on an internal model with undisclosed test-time compute, hindering independent reproduction and analysis by the community. Here we present a simple autonomous agent built on the publicly available GPT-5.5 Pro. The agent uses a problem-agnostic three-stage prompting pipeline: brainstorm, execute, and review. We ran eight independent trials with web search disabled. In each trial, the agent produced a correct or essentially correct disproof of the conjecture. The resulting proofs are diverse and do not merely reproduce earlier AI-generated proofs. Since GPT-5.5 Pro was released nearly a month before OpenAI published its disproof, our evaluation is free from data contamination. The system operates efficiently, using an average of 106.7 k reasoning tokens per trial. We release our code and all outputs, providing a publicly accessible AI agent to resolve this conjecture autonomously.

Authors

Institutions

Publication Details

Journal
Proceedings of the National Academy of Sciences
Published
2026-10-26
DOI
https://doi.org/10.1073/pnas.2621541123
Primary Topic
Artificial Intelligence Applications
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Autonomous disproofs of the Erdős unit distance conjecture with GPT-5.5 Pro

Yichen Huang
Proceedings of the National Academy of Sciences
Artificial Intelligence Applications
article

Autonomous disproofs of the Erdős unit distance conjecture with GPT-5.5 Pro

Yichen Huang
article en

Abstract

OpenAI's disproof of the Erdős unit distance conjecture marks a milestone in AI for mathematical research. However, the result relied on an internal model with undisclosed test-time compute, hindering independent reproduction and analysis by the community. Here we present a simple autonomous agent built on the publicly available GPT-5.5 Pro. The agent uses a problem-agnostic three-stage prompting pipeline: brainstorm, execute, and review. We ran eight independent trials with web search disabled. In each trial, the agent produced a correct or essentially correct disproof of the conjecture. The resulting proofs are diverse and do not merely reproduce earlier AI-generated proofs. Since GPT-5.5 Pro was released nearly a month before OpenAI published its disproof, our evaluation is free from data contamination. The system operates efficiently, using an average of 106.7 k reasoning tokens per trial. We release our code and all outputs, providing a publicly accessible AI agent to resolve this conjecture autonomously.

Proceedings of the National Academy of SciencesVol. 123(44)
College of the Canyons (US)
DeepMind, OpenAI
Openalex Percentile: Top 14%
Artificial Intelligence Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon