harness-engineering Harness Engineering (Part 3): What it costs, and what I measure Where the tokens actually go, which model belongs in which seat, why verification needs a budget, and what I measure now.
harness-engineering Harness Engineering (Part 2): Who checks the agent's work Thirteen bugs that survived testing, a review replay against 45 human-found bugs, and why the agent that wrote the code should never grade it.
harness-engineering Harness Engineering (Part 1): What it is, and why my CLAUDE.md wasn't enough What harness engineering is, from someone using an agent every day to ship a mobile app, and why rules in CLAUDE.md kept coming back until they became hooks.
