I Built My Own App and I'm Still Not Sure How

Everyone is talking about vibe coding. If you've spent any time on LinkedIn or X in the last few months, you've seen the posts. Someone built a SaaS in a weekend. Someone shipped a mobile app without writing a single line of code. Someone quit their job because they finally realised they didn't need a developer to build their idea.

I read enough of those posts that I decided to try it myself. What followed was not a weekend success story.

It was several weeks of confusion, accidental changes, a lot of agreeing to things I didn't fully understand, and somehow, at the end of it, a working fitness app on my phone that I actually built. Here's what that looked like.

The idea

I wanted a simple strength training app for myself. The ones I'd tried, Strong and Hevy, are decent but they always felt like more than I needed. The one thing I wanted that none of them did well was a recommendation for what weight to use next, based on how my last session actually went. Not a generic progression formula. Something that looked at what I completed, how hard it felt, and told me what to try next time.

I wrote a proper PRD, got the design direction sorted, and figured I was ready. I was not ready.

First stop: Kiro and Cursor

Cursor had been on my radar for a while and Kiro had just launched to a lot of buzz. Both are AI-powered coding environments. I fed in my PRD, my design references, and started working through the build.

For a little while it felt like forward motion. Then things started going sideways in ways I couldn't explain. A page I hadn't touched would change. Something I'd approved in one area would quietly break something else. And because I wasn't coming in as a developer, I was agreeing to suggestions I didn't fully understand, which meant I was also introducing problems I had no way of tracing back.

The tools weren't bad. They just assumed a level of technical comfort I didn't have. I was a product person trying to operate in a developer's environment and the gap showed pretty quickly. After spending enough time with both I stepped back and asked myself a more useful question: where should someone without a development background actually start?

Finding the right entry point

That brought me to Bolt, v0, and Replit. These sit at a different level entirely. Instead of dropping you into a code editor, they let you describe what you want and generate a working base from it. I fed in my PRD, design style, and references and the difference was pretty immediate.

Within a reasonable amount of time I had something that actually looked like an app. Not finished, not perfect, but a real foundation with actual screens and structure. Something I could look at and think, yes, this is the direction. v0 in particular worked well for me. Getting a solid base out of it was manageable in a way the other tools never felt at my level.

Taking it local

Once I had something I was happy with I downloaded the zip from v0 and moved everything onto my laptop. From there I worked locally, using Claude to help me edit. The workflow was fairly simple: open the relevant files, describe what was broken or what I wanted to change, and work through it piece by piece. UI fixes, logic issues, things that weren't behaving how I expected. Test, find the next problem, fix it, test again.

There was still a lot of back and forth. But seeing the changes happen directly in front of me and understanding why something wasn't working felt completely different from before. I was actually learning what I was building as I went, which wasn't something I expected to get out of this.

Supabase was a bigger lift than I'd anticipated. Getting the databases set up properly so all the different parts of the app communicated with each other took real time. I went in knowing nothing about how it worked. I came out the other side actually understanding it, and that felt like an unexpected bonus even if it slowed me down getting there.

The part nobody mentions

The app works. I'm still a little surprised that sentence is true, but it is. Aethon logs workouts, tracks previous performance, and recommends next weights based on how the session went. It does what I set out to build.

The part none of the weekend success story posts mentioned: to get the app onto TestFlight so I could actually use it on my own phone, I had to pay for an Apple Developer account. Not a huge cost, but an unexpected hurdle right at the finish line that nothing I'd read had prepared me for. Small thing, but exactly the kind of friction that never makes it into the highlight reel.

What I took from this

Vibe coding works, but it works differently depending on where you start and how comfortable you are navigating ambiguity in a technical environment. For someone without a development background, jumping straight into Cursor or Kiro is probably the wrong first move. Starting somewhere that generates a real base first and then moving into more detailed editing from there is a much more manageable path.

The other thing that stood out was how much having a proper PRD before starting actually mattered. When I knew exactly what I was building and why, the conversations with AI tools were sharper, the decisions were cleaner, and the output was better. When the brief got vague, so did everything else.

That last point is something I've been sitting with since finishing the build. More on that soon.