The Problem
Most fitness apps take months to build and six figures to launch. The market moves fast. By the time a traditional agency ships your MVP, the trend you were chasing has already peaked. I wanted to test whether AI-accelerated development had gotten good enough to compress that timeline to something that actually matched how ideas move — fast, iterative, and shipping before you’ve overthought it.
FitRush was the test case: a social fitness app where users could log workouts, track streaks, and share progress with friends. Simple enough to be buildable in a week. Specific enough to be real.
The Approach
I didn’t start with code. I started with the one-sentence value proposition and worked backwards. What does the user do in the first 60 seconds? What does session two look like? Where do they invite a friend? Those three questions defined the scope. Everything that didn’t answer those questions got cut.
The stack decision took about 20 minutes. React Native with Expo for cross-platform coverage, Supabase for auth and database, and a clean component library that wouldn’t require design from scratch. The goal was shipping, not perfection — and picking proven tools meant I wasn’t debugging infrastructure while also building features.
The Build
Day one was architecture and auth. Supabase handled user accounts, session management, and row-level security so I wasn’t writing auth logic by hand. Day two and three were the core workout logging flow — exercise selection, set tracking, and the data model that would power everything downstream. Day four was the social layer: following other users, the activity feed, streak tracking. Day five was UI polish and onboarding. Day six was TestFlight. Day seven was App Store submission.
AI did the heavy lifting on boilerplate — component scaffolding, API route generation, TypeScript type inference, and test cases. I directed, reviewed, and made architecture calls. The leverage wasn’t in replacing engineering judgment. It was in eliminating the parts of the job that don’t require it.
The Outcome
The app was in App Store review by end of day seven and approved within 48 hours. Full social workout tracking, streak logic, user profiles, and a working onboarding flow — from blank repo to live on iOS in one week. The development cost was a fraction of what a traditional sprint-based engagement would have run.
More importantly, it proved the methodology. This wasn’t a demo or a prototype. It was a shippable product built at a pace that matches the actual speed of ideas.
What I’d Do Differently
The streak logic had an edge case I didn’t catch until TestFlight — if a user logged two workouts on the same day, the streak counter miscounted. Small bug, but it showed up in the first round of testing. I’d build the core data integrity tests earlier in the week rather than treating testing as a day-six activity. The other thing I’d change: I underestimated how long App Store metadata prep would take. Screenshots, descriptions, keywords, age ratings — that’s a half-day of work on its own. I’d schedule it for day five and leave day six entirely for review prep.