Blog

Short, plain write-ups of things we built into the harness: what the problem was, what we tried, and what we landed on. Not philosophy; just the working notes, written like a blog post.

• How I trust what an AI builds — the whole net that keeps an AI's output trustworthy: tests, coverage, mutation, browser tests, a pattern library, and a phone-easy visual review. Once agents do the building, the human QA is the bottleneck, so most of it is automated and the bit that's left is made seamless.