Unhyped

AI cracked an 80-year maths problem

25 August 2026

Unhyped

AI just cracked a maths problem humans couldn't solve for eighty years. The reason it matters to your business has nothing to do with maths.

In 1946, the mathematician Paul Erdős asked a question that sounds like a puzzle from a kids' book. Scatter dots on a page, then arrange them as cleverly as you can. What's the biggest number of pairs you can get sitting exactly one centimetre apart?

For eighty years, the best mathematicians alive believed neat grid patterns were as good as it gets, and nobody could prove it either way. In May, an AI model found arrangements no human had thought of, better than the grids, and the proof passed checking by some of the best mathematicians in the world.

Most business owners wrote AI off over one rubbish email draft two years ago. You judged it on the worst thing it ever did for you, and while you weren't looking, the ceiling moved miles above where you're using it.

One caveat worth noticing. The AI produced the work, but humans checked every line before anyone believed it. That's the working relationship: it does the heavy lifting, and you stay the judge.

So dig out one task you decided AI couldn't handle, and hand it over again this week. The tool you tried back then doesn't exist anymore.