Summary
iOS indie developer Chris Raroque documents the full process of building an Apple Watch companion app for Amy — his AI-powered calorie tracker running at $3.4–3.5K MRR from 336 paying users at an 85% profit margin. The build used OpenAI’s Codex with a detailed prompt, simultaneously driving an Apple Watch simulator and an iPhone simulator, with the Xcode Build MCP server enabling Codex to compile, stream logs, and self-debug over roughly two hours of autonomous operation.
Real-world testing exposed a critical architectural flaw: the initial WatchConnectivity bridge between the Watch and iPhone proved unreliable, with Apple silently disconnecting the channel when the iPhone entered low-power mode. Raroque walks through his decision to port audio dictation and calorie calculation directly onto the Watch device, establishing a direct backend connection for latency-sensitive operations while retaining the iPhone bridge only for periodic data syncs — a change that eliminated stability failures entirely.
Beyond the technical build, the video offers candid business analysis: Watch adoption now accounts for 30–40% of Raroque’s own app usage and logging activity. He argues that Watch apps, Siri integrations, and home screen widgets may not individually move subscription retention metrics, but that stacking frictionless touchpoints creates the habit formation that combats churn in productivity apps. The video is a practical reference for solo developers considering watchOS development with AI coding agents.
📺 Source: Chris Raroque · Published September 08, 2026
🏷️ Format: Hands On Build
