A few months ago, Ralph had an idea for a word puzzle game. He opened Glitch and vibe-coded a prototype — fast, scrappy, proof-of-concept energy. It worked. It was fun. But it was a Glitch project.

That's where most side projects die.

This one didn't.

From prototype to Next.js

I took Ralph's working prototype and used it as the spec. Not a requirements doc — the running app itself. I vibed it into a real Next.js application deployed on Vercel. Same game logic, real infrastructure, production-ready stack. VWLS.xyz was live.

But we kept asking ourselves: could this be an app?

The PRD trick that changed everything

Here's where it gets interesting. Instead of starting a mobile project from scratch, we went back to the vibe-coded version and asked AI to describe the full state of the app — what it did, how it worked, every interaction. Then we turned that description into a proper Product Requirements Document.

That PRD became the blueprint for the iOS build.

We used Rork — an AI-powered mobile app builder — and fed it the PRD. What took weeks of traditional iOS development compressed into something we could actually iterate on fast.

The Apple wall

Getting an app approved by Apple is famously painful. We hit a hidden issue in the Apple Developer portal settings — one of those obscure configuration problems that doesn't have a clean error message, just silence and rejection.

We brought in Claude. Walked it through the dev portal, described what we were seeing, and it found the problem. Fixed, resubmitted, approved.

VWLS is live on the App Store

The app is out. A daily word pyramid game — you start with a vowel, build six words, 8 attempts to complete the pyramid.

Play free in the browser at vwls.xyz

Download VWLS Vowels on the App Store

What this actually means

A year ago, this path didn't exist. You'd need a mobile developer, an iOS engineer, someone who knew the App Store submission process inside out. Minimum 3–6 months, minimum $20–30K.

We did this in weeks with two people who aren't iOS developers.

That's the actual story. Not that AI is magic — it's that the barrier to shipping real software has collapsed. The question isn't "can we build this?" anymore. It's "what do we want to build?"

We're already thinking about the next one.