Writing, notes, and technical references I keep coming back to.
April 24, 2026
Lessons from shipping a regulated, customer facing LLM agent—and the lifecycle pipeline we built to make improvements measurable, repeatable, and defensible.
April 2, 2026
a coding agent is a stateful control loop, not a chat request. How I’d design the harness—loop, context pipeline, scheduling, and permissions—so it can run for hours without corrupting state.
March 23, 2026
More instances can make 429s worse when upstream quotas are global. How we reframed it as a coordination problem — queue, global limiter, and throughput shaping instead of scaling alone.