Vibe Coding All the Way to Production: A Google Play App Where AI Wrote 99.9% of the Code

A developer has published Yarrow AI, a native Android app, to Google Play, and 99.9% of its source code was written not by a human but by a neural network. This is not a throwaway weekend hack: the app is written in Kotlin and contains more than 50 screens, complex animation, multilingual support and an extensible architecture.

What kind of app did the AI write

Yarrow AI is a full-fledged native Android app published on Google Play in July 2026. According to the author of the case study, he wrote only 0.1% of the code by hand; AI generated everything else. This is not a prototype or a demo but a product that passed the store's review and is available to users.

The author separately stresses that the app is native: it is written in Kotlin, not built as a cross-platform wrapper around a web page. That detail matters, because native code spanning 50+ screens with complex animation is traditionally the case where AI generation breaks down most often.

• 99.9% of the source code was written by AI, not by a human
• The app is native, written in Kotlin — not a cross-platform wrapper
• More than 50 screens with complex animation
• Support for multiple languages and an extensible architecture
• Published on Google Play as Yarrow AI (com.yarrowai.yarrowapp)

"This is not just a 'quick weekend hack,'" the author writes in his article on

Habr.

Why the AI could handle this much

The AI coped with 50+ screens because the author prepared an extremely detailed specification in advance instead of asking the model to "build an app." That level of detail turned code generation from a creative task into a mechanical one.

The preparation had two key layers. First, a designer drew every screen of the app, so the model had a precise visual reference rather than an abstract verbal brief. Second, for every button the author specified exactly which APIs to call. With this much detail, the AI does not invent an architecture; it fills in a predefined framework, and the risk of hallucinations and rework drops sharply.

The case reflects the main shift in "vibe coding": the bottleneck moves from the code itself to the quality and completeness of the technical brief. The more thoroughly each screen and each action is described, the fewer fixes are needed after generation. In effect, the human's role shifts from the person who types the code to the person who designs the system and states requirements so that the model has no room for guesswork.

Was it worth it

The author asks outright whether the method pays off and suggests looking at the numbers rather than impressions. Writing a detailed specification for 50+ screens, with every screen drawn out, is a large job in its own right. It falls entirely on the human and comes before any generation.

"Was it worth it?

Let's look at the numbers," the author writes in his Habr article.

The case's main takeaway is not that AI "replaced" the programmer, but that under strict engineering discipline, AI generation scales to a real product in an app store, and not just to a one-screen weekend demo.

What this means

The Yarrow AI case shows that "vibe coding" has grown up enough for production, but not on its own: behind the 99.9% AI-written code stands 100% human work on the specification, the screen design and the API descriptions. AI takes over the routine of writing code, while the value shifts to whoever can define a task clearly and completely.

Frequently asked questions

How many screens does the Yarrow AI app have?

More than 50 screens, with complex animation and support for multiple languages.

How was almost all of the code written by AI?

Thanks to a very detailed specification: a designer drew every screen, and for every button the author spelled out exactly which APIs to call.