AI Board
A deliverable-centric AI workspace: define what "good" looks like, get an editable workflow to produce it, set review criteria, then run it and receive scored results. Review is a first-class activity, not an afterthought.
AI Board inverts the usual chat-to-output pattern. Instead of prompting and hoping, you start from the deliverable: define its shape, its dependencies, and your definition of good. The system proposes a workflow to produce it — which you can edit — and you attach review criteria before anything runs.
When the run finishes, you don't just get output; you get output scored against the criteria you set. That makes reviewing AI work a structured activity instead of a vibe check.
Under the hood it's the standard stack I use for most of my apps: React Router in framework mode, Drizzle + Postgres, better-auth with Google OAuth, and the Anthropic Messages API for all AI operations.