# @alanou — 2025-08-29

♥0 ↻0 · https://x.com/alanou/status/1961469613499129970

@tessera_antra I say LLMs are human-shaped. They are trained to generate data that looks like it was generated by humans. I suspect the easiest way to do that is to mimic similar cognitive structures. They are piecewise linear approximations of a gestalt human mind. ReLU is a linear piece.

tags: author:alanou, kind:tweet, thread-context, year:2025