Which is great, because there's some problems LLMs aren't well-suited for but other computational tools are, and astonishingly LLMs can sometimes detect when those other tools should be used. I am waiting for somebody to combine an LLM with a theorem proving system.