Guides · 3
How the AI here works.
Concrete write-ups on the chat panel that runs this site — how it retrieves and grounds its answers, and how every answer gets graded. Built from the real pipeline, with real numbers.
Interface
Generative UI: when the chatbot shows instead of tells
The chat doesn't just write prose — it renders real UI. How four tools fire on overview intent, the rules that keep them from over-firing, and how it all wires up.
June 2, 20266 min read
Generative UIAI SDKAI
Retrieval
How the chat works: RAG, grounded in my own writing
The RAG pipeline behind this site's chat — how content becomes chunks and embeddings, and how the model answers only from what it retrieves.
June 1, 20267 min read
RAGEmbeddingsAI
Eval discipline
LLM-as-judge: a regression caught, a regression fixed
How an LLM-as-judge eval suite caught a same-day regression on this site's chat panel — what broke, why, the one-line fix, and the result on the rerun.
May 25, 20265 min read
EvalsRAGAI