10 months ago this was a competitor list. Here's what it turned into.

Ten months ago the whole product was one thing. You typed an idea, we showed you who was already doing it. That was the first launch. It got 15 upvotes.

What changed since:

v2, April. We stopped letting the model guess. Every analysis now runs on real data: Reddit threads, app store reviews, PH launches, search demand. The output became a PRD you could hand to someone, plus a first set of mockups. Still one big bundle though. You got everything or nothing.

v3, August. We split the bundle into steps that read each other. Idea, strategy, competitors, PRD, IA, wireframes, design, prototype. The IA comes out of the PRD, the wireframes come out of the IA, the design reads the wireframes. The IA is a tree you can edit, with the wireframe sitting under each screen.

Design, the step we skipped for two versions. You pick a style from 250+ real apps and the sheet comes with the colors and type rules already pulled out. Then the wireframes get dressed in it. Our editor is not Figma and will not be. That is why Figma export sends real nodes, not a PNG.

Teardown. Paste an App Store link or drop screenshots. We watched 8 screenshots turn into 17 screens and 58 named components. It is for reference work, not copying. Most people already have an idea and want to see where it sits next to the apps they know.

MCP. Connect Claude Code or Cursor and it reads the PRD, IA, wireframes and style guide straight from the project. No more pasting a doc into a chat window and hoping it stays in context.

Still rough: the prototype is HTML, not native code. And the pipeline is only as good as the idea you give it. Vague in, vague out.

One question for people who build with AI tools. When you start, what do you feed it first: a prompt, a document, or screenshots of an app you like?