The studio started in 2026 with a simple observation. AI models can now do a large share of the work that goes into software, from research to code to review. What they can't do on their own is stay pointed at the right goal, or judge their own output honestly. That part still needs a system, and people who decide.
So we built one. ARCS, our intelligence system, splits work into bounded tasks, hands each task to a specialist council and has every result checked by a model that didn't make it. Every task carries the long-term goal it serves: its north star.
We use the system in two ways. We build and run our own products with it, five so far, from a rain-advice app to a learning platform for teams. And we put it to work for companies: new products, internal tools, and reviews of code that has to hold up.