The motivation
By nature, and in my work, I'm a team player and team leader. But sometimes I get depressed from constantly dealing with people problems, interpersonal relationships, waiting for someone, reporting to someone, etc.
And now we have AI (in my case Claude Code). I finally got to try whether I can build something on my own – without programmers, designers, without constantly waiting for someone or something.
My background: I programmed for years, but for the last 10-12 years I've been primarily a Product Owner with a strong UX focus. And this is where my opportunity comes from: I can communicate well, I can write good specs, focus on MVP and context. At the same time, I wanted to dive deeper into AI because it's a huge opportunity today – maybe a once-in-a-century kind of opportunity.
The project
I decided to build a cross-platform app in React Native using Claude Code – simple enough that I wouldn't need a backend for now, with some real value, and most importantly, I'd learn how to work with Claude Code.
I chose to build a simple subscription tracking app (primarily for digital services like Netflix, Spotify... you know the drill).
I didn't create this app to make money (although you can buy it), but because if I can pull this off, I can tackle much more interesting business projects.
The process
From the start, I knew that building an app solo requires iteration in small chunks – just like with people, AI works better with smaller pieces and maintains context better. But it's good to know the full scope.
Here's how I approached it:
a) Preparation: I prepared requirements in the form of User Stories, fairly high-level. Writing the initial specs took about 2 hours, then I iterated with AI for another half day until we had around 20 basic User Stories.
b) Technical architecture: I iterated on the technical solution with AI (since I have a programming background and work with developers daily, this simplified things significantly). In total, preparing requirements and app architecture took about 1 day.
c) First surprise: I wanted Claude Code to work in small iterations as I defined them, but we didn't quite understand each other – the first iteration basically built the entire app core. It took about 30 minutes. Surprise – the core was functional. I was excited but also conflicted, because I wanted to test piece by piece. It was clear I needed to proceed slowly and carefully from here.
d) The framework I developed:
I started refining the app screen by screen, User Story by Story, and gradually built the following workflow:
Timeline
You're probably wondering how long it took to build the app. I started in mid-November and finished in early March, working about 2 evenings per week and every other weekend. So quite a bit of work, relatively speaking.
What I learned
Check it out
It was a great experience and I'm already looking forward to the next one – a much bigger project this time.
If you're curious what the app looks like and how it all turned out, check out https://moonsubs.com or download it from the https://apps.apple.com/app/id6756514083.
I'd love to answer questions about building with AI (I know this article is a bit brief) and I'd appreciate any feedback on the app.
Thanks for reading and have a great day!