- The author wanted a habit tracker app with specific features like a long list, local storage, custom emoji icons, flexible goals, notes, and reminders, but existing apps like Streaks, HabitKit, and Grit lacked some features or were too expensive.
- Using AI (Cursor) to build the app in Swift, the author created a working prototype in 6 hours, but the code was messy—views were 1000 lines long, inconsistent, and full of bugs like iCloud sync not working and performance issues.
- Manual refactoring fixed the code structure, but new problems emerged with iOS 26 updates that AI couldn't handle, requiring months of debugging and cleaning up the codebase.
- The final app, HabitTed, was released on the App Store after a year, free for up to 6 habits with a one-time purchase option, but the experience highlighted the 80/20 rule: AI gets you 80% there, but the last 20% takes 80% of the time.
- The author reflects on the risks of relying too heavily on AI—losing coding skills, burnout among developers, and a potential shortage of competent developers if the AI bubble bursts.