Field notes on product strategy, delivery, launch work, AI tooling, and the systems behind teams that outgrew spreadsheets.
RAG is just looking things up before answering. Here is what chunking, embeddings and retrieval actually do, why fine-tuning solves a different problem, and the five ways RAG breaks in production.
Agent harness, loop engineering, LLMOps, and eval sound like heavy jargon, but each one is genuinely simple. Here's how these four building blocks turn a raw LLM into an agent system you can actually trust.