Founder-led, AI-native: how a small studio ships

ONLYAPPS is small on purpose. It has a founder, and it has a system of AI agents.

I wanted to find out whether that is enough to make apps that feel finished. Not impressive on a slide, but pleasant on a Tuesday when you just want something to work. This post explains the arrangement, because I would rather you hear it from me than guess.

The split. I decide what we build and whether it is good enough. The agents do the research, planning, design, building and testing, and they draft replies to support messages. They work inside rules, and the rules matter more than the agents.

The rules.

• Whatever an agent makes is reviewed by a different agent. Work that is not good enough goes back.
• Spending money, publishing an app and anything that cannot be undone need my explicit approval. The system stops and waits for me.
• Every screenshot on our site is a screenshot of the real app. Animated previews are labeled Preview.

What this does not mean. It does not mean nobody is accountable. A person decides what ships. It does not mean the agents never make mistakes. They do, which is why there is a review step and an approval step.

Why bother. A small studio has limited attention. This setup lets me spend mine on the question that decides whether an app is worth installing: is it good? Does it do one job, explain itself and respect your data?

Our first two apps, Cast to TV and FitGIF, are coming to Google Play. If something is unclear or wrong, write to us at support@onlyapps.work.

— James Jung