This will massively solve, the mathematical proof review process bottleneck.
because for humans this process is so hard as normal mathematical proof leaves thousands of steps unstated:
pro mathematicians can infer them, but the Lean system cannot, so every definition, lemma, dependency, and tiny logical step has to be written explicitly and connected to formal libraries that are still incomplete.

So in this of Fermat's Last Theorem, that meant converting a huge modern proof into tens of thousands of machine-checkable pieces, and this work was expected to take years by humans.

But, Claude compressed that years-scale formalization task into 11 days

So now, AI can make formal verification practical for much more mathematics. And this will be so important because we are already seeing that AI increasingly is producing far more proofs than humans can manually review.