How I built Kujua Focus with AI, from research brief to my phone

Kujua Focus is a digital wellbeing app built around one idea: a friend holds the key. This is the exact order I built it in, so you can copy the process for your own idea.

Phase 1: Think before you build

I started with a research brief covering the problem, competitors and the habit science behind it. The AI turned that into a PRD, then a feature backlog tagged MVP, 1.1 and 2.0.

Phase 2: Design before you code

I wrote the visual style once, as a design.json file, then built a single demo screen that used every component. I checked it on my phone with Expo Go before building anything else.

Phase 3: Build in layers

1. 01Every screen prototyped with fake data: onboarding, the urge-moment shield, the Spotter, streaks
2. 02Fake data swapped for real data, so new users get a real day one
3. 03The hard native part: an Android module that actually blocks other apps
4. 04AI added one ability at a time, running on a server so the API key never sits in the app

Phase 4: Ship it to a real device

Every working step ended in a Git save point and a push. Then a cloud build turned the project into an install file for my phone. The first build failed. The disk was full.

bash$ eas build --profile development
✗ Build failed: the disk was full
$ npm cache clean --force   # freed 2.7 GB
$ eas build --profile development
✓ Android install file ready

The loop inside every step

Ask, build, try it on the phone, react, save, update the docs. Twelve steps sounds like a lot. Each one is just that loop until it works.

Written by

Samuel Audu

I build real products with AI and teach the process in a live class. More about me