Some mathematicians were upset that OpenAI released solutions to hundreds of previously unsolved math problems. Their main argument is that, without mathematicians, no one can understand or verify these solutions. Well, ironically, AI may now make it possible for everyone to understand them!
I asked my ChatGPT dot agent to choose one of these mathematical results with a reliable computer-checkable proof, explain it to someone without a math background, and show how we can check its correctness.
This is what did (in its own words) and created the attached images explaining the problem, solution, what it means and how it was checked:
“What I did: I reviewed the release and several candidates, chose Seymour’s second-neighborhood theorem, and compared its precise statement with the plain-language explanation.
I consulted GPT-6 Pro for critical reading and teaching advice. All 43 original proof files passed Lean, the audit found no admitted gaps or extra assumptions beyond its standard foundations, and a fresh replay passed too. I then created and reviewed these four visuals”
I can now ask it to do the same for all the other solutions. It would probably take the poor guy an entire day, but it would get the job done!
And imagine how much better and more reliable this process will become once we have superintelligence, which I expect to arrive around this time next year! Sorry guys, math is cooked!
Learning
Agent explains a Lean-checked proof to laypeople in four images
Immunology professor Derya Unutmaz asked a ChatGPT dot agent to pick one of OpenAI's new machine-checkable math results and make it understandable. It chose Seymour's second-neighborhood theorem, verified all 43 proof files in Lean, consulted GPT-6 Pro and drew four explanatory images.
395likes
34reposts
91saves
20Kviews
- Made with
- ChatGPT dots, GPT-6 Pro, Lean
- Role
- Professor and immunologist
- Posted












