LLMs / 2026-01-30

Note: An LLM was not used in writing this microblog entry.

This entry is part of a microblog series called LLMs


A paper discusses a similar topic that I’ve addressed in this microblog, which is the negative relationship between skill and using GenAI.

I haven’t generated anything since the last update. But I hand wrote a little experimental library that uses a free applicative to both declare a GBNF grammar and consume the output back into Haskell. I didn’t add an Alternative instance (yet) to support disjunction in the grammar, but this direction appeals to me. I’ve been dabbling with the same llama3.2 (3b model) via llama.cpp’s server locally on the same MacBook Pro M4 Max. It’s quite easy to get reliable outputs. I intend to dabble more with this in coming weeks.

I’ve been recently thinking that the whole phrasing around interacting with LLMs has become increasingly anthropomorphic, and thought of a clever mental hack to avoid it: instead of “talking to Claude,” say “talking to the computer,” (doesn’t that sound silly!) and instead of “AI generated” or “LLM generated,” simply say “computer generated.” (How dull!) I find it thoroughly cuts through the illusion that we are doing anything more. But YMMV.