Skip to content

Notable cases of AI contributions to Erdős problems

Nat Sothanaphan edited this page Jun 30, 2026 · 8 revisions

The wiki is no longer updated. The latest data is as of Jun 30, 2026.


We record some notable cases of AI contributions to Erdős problems.

We hope that it is useful to understand AI trajectories and to gather preliminary information for further investigation.

Problem [1026]

Timeframe. Dec 2025.

Significance. Showcased various forms of AI assistance and usefulness in a mathematical collaboration.

Final outcome. Solution with strengthened conclusion.

Resources. This blog post.

Problem [728]

Timeframe. Jan 2026.

Significance. Regarded as the first autonomous and nontrivial solution to an Erdős problem by AI systems.

Final outcome. General result giving solutions to this problem, [729], and [401].

Resources. This arXiv preprint. This blog post.

Problem [1196]

Timeframe. Apr 2026.

Significance. Obtained solution to a problem that human experts have already spent substantial effort on.

Final outcome. Novel technique resolving this problem, [1217], [164], and two other conjectures. Inspired solution to [858].

Resources. This arXiv preprint. This blog post.

Problem [90]

Timeframe. May 2026.

Significance. Resolved major open problem in a mathematical area with prior widespread expert attention.

Final outcome. Also disproved [92]. Admissible exponent recorded as a new optimization constant. Inspired this arXiv preprint on [52].

Resources. This arXiv preprint. This arXiv preprint. This MathOverflow question. This blog post.

Clone this wiki locally