We are trying something specific at Team AM, and we want to write down what it is actually like as we go — not a highlight reel, but the real trajectory, including the parts that do not work.

The bet

Let AI agents do the production work — the code, the design, the store listings, the endless busywork — while a human owns the quality gate. One person, working with agents, can run a whole portfolio of apps that way. The interesting question is not whether it can produce things. It can. The interesting question is what the human’s job becomes when it does.

What we shipped

We spent this stretch getting a set of niche maintenance apps ready for the App Store. One is in review; the rest are queued behind it. An AI agent did essentially all of the listing production — marketing screenshots for iPhone and iPad, the store copy, pricing, compliance, the App Store Connect plumbing — and we reviewed and approved each step.

The division of labor that emerged is the whole story: the agent produces; we decide what to ship and what to reject.

The honest lesson

When AI makes building nearly free, building stops being the hard part. Distribution does.

Our one app that is already live gets a trickle of installs a day with no marketing, and no amount of extra polish changes that. The constraint was never the product — it is being found. So the next stretch is about learning distribution in the open: store optimization, ratings, and a careful look at whether paid acquisition can ever pay for itself at our prices. (Early answer: not yet.)

A failure worth sharing

The useful posts are the ones about what broke. So: the agent confidently produced a page that rendered completely blank. It looked finished. It was not. A human noticed, and it got fixed in a few minutes.

That is the model working exactly as intended — and it is also the entire risk in one sentence. When production is cheap and everything looks done, the quality gate is the whole job. Rubber-stamp it and the mistakes ship invisibly. The scarce thing is no longer the ability to make things; it is the judgment about which ones are actually good.

What’s next

Distribution experiments on the live app, and learning the path to selling software to small operators — marinas, property managers, small fleets — rather than chasing big companies. We will report real, relative numbers as they move.

More soon. We will keep these honest.