In the past we would sometimes forgive even very bad writing if a paper contained a correct proof of an important, interesting result. But we probably should not do that anymore. An important job of the human researcher in posession of a proof machine is to digest the proof and
Aaron Roth@Aaroth
What are our jobs as researchers now that we have proof machines? The raw proofs that come from LLMs are difficult to understand, even if correct. So its now easy to quickly write many badly written papers that nevertheless contain correct proofs of interesting results.
6:28 AM · Jul 1, 2026 · 10 Views
