Things I've written down.
Ruby, Rails, performance, security and applied machine learning. Updated when something feels finished.
Temperature is not a creativity dial
Temperature, top-k and top-p are three different operations on one probability distribution. Treat them as vibes and you'll tune the wrong one. Watch them act on a model small enough to see.
Your RAG demo is lying to you
Every shortcut you take to make retrieval cheap — a smaller model, an approximation, a reduced dimension — can invent failures the real system doesn't have. Check the behaviour you're about to teach or ship.
pgvector or a dedicated vector database?
For most applications the answer is the database you already run, and the threshold where that stops being true is higher than the marketing suggests.
Chunking is a retrieval decision, not preprocessing
How you split documents determines what is findable. Everything downstream inherits that choice, and almost nobody measures it.
Your RAG chatbot doesn't have a model problem
It has a measurement problem. Why swapping the model is the last thing to try, what to instrument first, and how an aggregate score hides the failure that matters.