AI coding agents make building cheap. Taste still matters.
When AI coding agents make implementation cheap, engineering judgment decides what is worth building and maintaining.
HomeWriting
Writing
Notes on LLM inference, agent memory, and building useful software with AI coding agents.
Agent memory after RAG: search, revise, forget
Why long-running AI agents need iterative search, revision, and forgetting instead of a fixed RAG retrieval step.
Serving LLMs is a product decision
How to weigh quality, latency, GPU utilization, and cost when choosing between self-hosted LLM inference and model APIs.