Tao: 'solved' isn't the same as 'advanced'

Mathematics advances through the route, not the endpoint, and the route is what trains the next generation. Terence Tao, a Fields medalist and one of the most cited living mathematicians, is arguing that AI is collapsing the route at scale.

The popular reading is whether AI is 'good at math' or 'bad at math.' That is the wrong axis. A tractable open problem is not just a thing to be answered; it is the scaffold that produces the methods, intuitions, and trained mathematicians the next decade depends on. Flatten such a problem with machine effort before it does that work, and the field accumulates solutions while losing the path that produced them.

Tao's framing, drawn from his March preprint with Tanya Klowden, is that this stock is now being 'mined in a non-renewable fashion.' His mechanism is specific: even the rumor of someone working on a problem can trigger a massive AI effort to flatten it before the original research project reaches its full potential. The next paper still gets written. The training ground underneath it does not.

There is a clean falsifier. If AI-generated solutions produce new tractable problems as fast as they consume old ones, the non-renewable framing fails. Tao's counter, in the preprint, is that mathematics is a human practice, not just truth-storage: a Lean-verified file without a vouched argument is socially thin inside the field. The route is the unit of progress, and the route does not survive being skipped.