I'm Not a Developer. I Shipped 3 Apps to the App Store This Year Using AI.
I'm a product designer. I can't write Dart from scratch. I don't know what Riverpod is. But I shipped 3 apps to the App Store this year, and two of them make money.
Here's what I learned.
The vibe coding trap
When I started using Cursor to build apps, I'd describe what I wanted and let it generate everything. The first version always looked great. The second feature broke the first one. By the third, I was spending more time debugging than building.
The problem wasn't Cursor. It was the foundation. AI generates code in isolation — each prompt creates new patterns, new file structures, new conventions. After 50 prompts, you have 50 different approaches in one codebase.
The structure matters more than the code
The breakthrough came when I stopped letting AI decide the architecture. I started with a production scaffold — proper folder structure, proper state management, proper patterns — and then used AI only for building features on top of it.
The AI went from "let me create a new folder structure for this feature" to "I see the existing pattern in the profile feature, I'll follow the same structure."
Night and day difference.
What actually works for non-developers
- **Start with scaffolding, not a blank file.** Use Shipeed or similar tools to generate a production structure. Don't let AI decide your architecture.
2. **Use AI guide files.** CLAUDE.md, .cursorrules — these tell your AI how the project works. Without them, every prompt starts from zero.
3. **Build one feature at a time.** Don't describe your entire app in one prompt. Get the scaffold, then add features one by one. "Add a profile page that follows the same pattern as the home page."
4. **Don't fight the patterns.** If the scaffold uses Riverpod, don't ask AI to use Provider instead. If it uses GoRouter, don't ask for Navigator.push. Consistency is everything.
5. **Ship the MVP.** Your first version doesn't need 20 screens. The scaffold gives you onboarding, home, settings. Add 2-3 features and submit to the store. You can always add more.
The results
My expense tracker took 2 weeks from idea to App Store. Revenue: $340/month from a $2.99 subscription. My habit tracker took 3 weeks. Revenue: $180/month.
I'm not a developer. I'm a builder who uses the right tools. The code doesn't have to be perfect — it has to be structured well enough that AI can maintain it.