Where it comes from
All four came out of a real project: an iPhone app built by a team of Claude sessions, taken from its first decision to a beta with testers in 9 days. The owner wrote no app code and made every decision.
Why I built this
I wanted to build an iPhone app without it being limited by what I know. I'm not a Swift developer, and I didn't want to learn how to structure a Swift project before I could start. So I let Claude write all of the code, and I kept every decision for myself.
That worked, but problems came quickly. The biggest one was context. The longer a session ran, the faster it burned through my usage, and most of what it carried wasn't needed. But if I started a fresh session, I had to explain everything again.
Having a chief of staff also made the whole setup modular. When I need something new done, I add a session for that one job and connect it to the chief of staff, without explaining the project to it. Once the app was in beta, I added a session that reads the testers' feedback, and a tester that drives the iPhone Simulator and tries edge cases I'd never have thought of.
kl-autonomous
12blocking problems the reviewer caught before anything merged
12restarts, with nothing lost
18 / 30desk items closed on the desk's first day