About
A small studio with one narrow thesis.
AI can write code faster than any team can review it. That gap is the problem we work on. Ai Pro builds the missing layer: specifications that are precise enough to check, and workflows that keep an agent inside them.
What we believe
A specification that cannot fail a test is not a specification — it is a wish. Most documentation decays because nothing forces it to stay true. We treat documentation as an executable artifact, and the test suite as the proof that it still matches reality.
The same applies to AI output. An agent is only useful if the work it produces can be reviewed, bounded, and rolled back. That requires more than a good prompt — it requires a charter that says what the product is for, decisions recorded with their reasons, and hard limits on what may change.
What we build
- DrawSI — turns architecture diagrams into verifiable technical specifications, so the drawing and the document cannot drift apart.
- Agent workflows — multi-agent setups with a governance layer: charter, decision records, scope guards, and traceable changes.
- Developer tooling — TypeScript, monorepo build graphs, browser automation, and Chrome extensions built for production teams.
How we work with clients
We start by writing down what success means in numbers, and what is out of scope. Then we specify the change, build it, and prove it with tests. If a request does not serve a stated goal, we say so before starting rather than after.
We are a small team by choice. That means direct communication with the people doing the work, and no account layer in between.