The entire app was built through vibe coding with Claude Opus using Claude Code - describing what I wanted and iterating on the implementation together. It features zone-based organization, CloudKit sync, and both "simple" and "advanced" modes to accommodate different cognitive loads.

Still squashing bugs before App Store submission, but wanted to share early and get feedback from the community. The app uses a gentle, non-judgmental approach to organizing - no before/after shame, just progress tracking and small wins.

Tech stack: SwiftUI, Core Data + CloudKit, MVVM + Coordinators, Semantic Colors

I had pretty decent success, so far - over the course of nearly a month, Claude Code only had a handful of instances of context poisoning and it felt like it got things right about 80% of the time. It was a daily experience of handholding and guiding, however. Still better not to take ones eyes off of the code, as I caught at least a couple of mistakes per day that would have led to a divergence from the expectation. Overall, very pleased with the experience, even if left with a slight headache.

There were some neat aspects to using the LLM, such as it creating semantic colors appropriately (after some trial and error) and using it to generate dynamic "smart behavior" without the app itself needing to reach out to an LLM to get feedback.

Would love feedback on the concept, UX approach, or the experience of vibe coding production iOS/macOS/tvOS apps with AI

TestFlight: https://testflight.apple.com/join/Kj9MZ9Mx

Website: https://roomcycle.app