avatar
Mike Wasson @mikewasson.net

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.

aug 29, 2025, 12:33 pm • 1 0

Replies

avatar
James Felix Black @tft.io

I often wonder if the invariants of a strong, rich type system like Haskell’s could outweigh the relative paucity of Haskell code in the wild and make it possible to get good results out of the models.

aug 29, 2025, 12:51 pm • 1 0 • view