Workwright
A protocol for humans and machines to build things together.
tl:dr;
Brief → Defense → Crit. Three words for the whole thing.
Workwright is a protocol where every change — whether it comes from a human or an LLM — goes through the same loop: propose a brief, defend the choices conceptually, submit to crit. Trust is a gradient earned through accepted work, not a permission level assigned by an admin.
The insight: human-in-the-loop fails not because humans are slow, but because the loop is too big. Make the unit of work small enough that a human can have a taste reaction in under a minute, and suddenly the loop works.
Built in one night. Wrights (LLM agents), humans, and teams all use the same CLI, the same protocol. The system doesn’t care what you are. It cares whether your work holds up under crit.