I've Built AI Agents. Now I'm Learning to Build Them Properly - From the Ground Up.
Starting a build-in-public series on agent engineering - the craft, not just the code.
Jun 18, 20263 min read64

Search for a command to run...

Series
A build-in-public series on agent engineering. Each lab produces working code, passing tests, and an honest write-up of what broke. Building Conductor - an AI agent for data integration - from scratch.
Starting a build-in-public series on agent engineering - the craft, not just the code.

Eval-first agent development: how writing tests before code changed every product decision I made.

Building the minimal ReAct loop that separates structural failures from behavioral ones — and the three tool design rules that show up before the lab is over.
