hallucinate.cc

This is a small personal corner of the internet where I write up things I learn while building with large language models — agent loops, tool use, browser automation, retrieval-augmented generation, and the very real failure modes that come with all of it.

The name is a tongue-in-cheek nod to the most famous LLM bug: hallucinations. A lot of what gets posted here is about reducing them in practice — better prompting, grounding, evaluation harnesses, and knowing when to tell a model to just say "I don't know."