I created a little RAG chatbot connected to some old school D&D rules (Old School Essentials / BX). It runs completely offline using llama3. I ran into some interesting issues, so I've been trying to document as I go in case I turn it into a tutorial.

I'm starting to feel more and more like half the AI terms thrown around are just the same thing re-marketed 😐

I bet someone's in trouble today at The Atlantic 😬

AI is great at filling in code scaffolds, but it struggles with vague, broad goals.

So when I write test descriptions like this, it has context. AI can infer function names, required params, and even edge cases.