Writing
Field notes
Practical notes on shipping production AI features, scaling systems, and getting real leverage out of AI-assisted development.
-
MCP Went Stateless: A Migration Checklist for Servers
The 2026-07-28 MCP spec removes sessions and the initialize handshake. Here is what to change in a server you run today, and what can wait.
-
Choose Boring Technology on Purpose: The 0-to-1 Stack
At the start of a product the goal is learning whether anyone wants it. That goal should pick your stack, and it points to boring technology almost every time.
-
Rolling Out AI Coding Tools Without Wrecking Code Quality
AI coding assistants speed a team up only if review and testing keep pace. Here is how to roll them out without a pile of PRs nobody understands.
-
Why Your RAG System Returns Garbage (and Where to Look)
When a RAG system gives bad answers, most teams rewrite the prompt first. The prompt is rarely the problem. Here is the order to debug it in.
-
Build vs. Buy: A Framework for Your First AI Feature
Most teams start build-vs-buy with the technology. Start with the cost of a wrong answer instead, then decide what to build, buy or skip.
-
2.9s to 300ms: A Systematic Approach to API Latency
A repeatable way to cut an API's p95 latency by an order of magnitude: find where the time goes, then fix the biggest block first.
-
Building an Autonomous Coding-Agent Fleet
What it takes to go from one AI coding assistant to a fleet of agents that ship real work: orchestration, isolation, and the verification layer.