One internal non-writing team runs setup on their real voice, as simulated external users.
Instrument:
- time-to-working-setup (target < 30 min, one sitting)
- completion rate
- every question they asked a human (each = a docs gap → its own issue)
- before/after quality delta on their own copy (Phase 4 output)
Kill/iterate condition from the proposal: if they can't finish in one sitting or the demo doesn't visibly improve output, the protocol is wrong — not the user.
🤖 Seeded by @daveh-beep's agent while standing up the repo.
One internal non-writing team runs setup on their real voice, as simulated external users.
Instrument:
Kill/iterate condition from the proposal: if they can't finish in one sitting or the demo doesn't visibly improve output, the protocol is wrong — not the user.
🤖 Seeded by @daveh-beep's agent while standing up the repo.