shameless self-plug: two hard things in computer science https://blog.est.im/2026/stderr-04 https://blog.est.im/2026/stderr-04 I think LLMs are reall good at

shameless self-plug: two hard things in computer science https://blog.est.im/2026/stderr-04 https://blog.est.im/2026/stderr-04 I think LLMs are reall good at conjecturing based on existing knowledge, but inventing new tools/lemma and new paradigm? Not really.