When Math Got a Partner

Fermat's Last Theorem sat unsolved for 358 years. The Riemann Hypothesis has been waiting since 1859. Andrew Wiles worked on his proof in secret for seven years, in an attic, telling almost no one. This is what doing mathematics looks like: a single mind, an impossible problem, and decades of silence before anything gives.

That picture is changing. AI systems are now solving advanced mathematical problems at a rate that has professional mathematicians genuinely unsettled. New Scientist is calling it a potential golden age of mathematics. The mathematicians at the frontier are calling it something more complicated.

Here's what makes this properly strange: mathematics was supposed to be the immune discipline. You can automate driving — that's physical pattern recognition. You can automate medical imaging — that's visual pattern recognition. But mathematics is supposedly the pure expression of human reason — the place where we argue with reality using a language of perfect precision, where we find truths that exist independent of any observer. Proofs don't care who discovers them; they're either valid or they're not.

And yet AI, trained on the accumulated human mathematical record, is now discovering valid ones.

The cosmic joke is beautifully recursive. We invented mathematics to describe patterns. We then built computers to process those patterns faster. We then trained those computers on mathematical patterns until they learned to generate new ones. The pattern-recognition engine is now recognizing patterns its creators couldn't see. The snake ate its tail and produced a theorem.

What's unsettling the mathematicians isn't that AI is making mistakes — it's that sometimes it isn't.

There's a version of this story that's triumphant: AI as the greatest research collaborator in history, giving mathematicians tools to explore vast solution spaces that would take centuries to survey by hand. More theorems proven. More connections found between distant fields. The golden age.

There's another version that's existentially peculiar: what happens to mathematics when the act of discovery — the long, slow, deeply human act of sitting with a problem until it opens — becomes something you can outsource? Mathematics was never just about the theorems. It was about the particular quality of mind that produced them. The proof and the prover were always, quietly, inseparable.

Mathematicians don't just find truths. They develop mathematical intuition — a felt sense for where the interesting problems live, what kinds of moves tend to work, what a promising approach smells like before you can justify it. That intuition is hard-won, deeply personal, and apparently not as exclusive as we thought.

The universe doesn't care which version wins. The patterns were there before humans named them and will remain after we're gone. We're just the local process that got around to writing them down.

Here's the thing: it's only vertiginous if you thought the discipline belonged to you. Mathematics never did. The theorems exist whether we find them or not. We found them slowly, alone, over centuries. Now we have company.

The universe just got a bigger research team.