Beto: My first iOS app with ~40 daily active users

I run without music. I wanted to hear my breath and footsteps, but I kept drifting from a target cadence five minutes in. The obvious fix was a metronome I could set to 170 SPM and forget.

Most alternatives I found were musician-first, crowded with controls or gated behind subscriptions. Running is the wrong moment to manage software. I needed one number, one action and a beat I could still trust after the phone went back into my pocket.

The most important design-system decision was not making one. I used Apple’s native components, semantic colours, materials and familiar control patterns, then spent custom design effort only on the beat, cadence and cross-device experience.

That was a speed decision and a product decision. Familiarity reduces attention cost, which matters more when someone is moving than when they are sitting at a desk.

Onboarding asks one useful question—how do you run?—and turns the answer into a sensible starting cadence. It avoids a tutorial while still preventing a blank, generic first run.

Free Run owns the common path. Presets handle warm-up, tempo and cooldown segments without turning the default experience into a workout builder. Empty, signed-out and configured states were designed as first-class parts of the product rather than cleanup.

The Watch is not a miniaturised phone. It prioritises glanceable cadence, elapsed time, large transport controls and precise wrist adjustments. Playback is independent, so either device can own a run; durable settings, presets, language and idle cadence stay in sync.

Cursor accelerated implementation; it did not decide the product. I worked in the same loop I use with engineering teams, only the handoff happened inside the same person.

The reliability work became more important than another screen. A shared, sample-addressed audio clock derives each beat from the audio engine frame counter instead of a wall-clock timer. The test suite covers 30-minute runs across 101 tempos, three sounds and seven accent modes: 2,121 combinations checking every onset and cumulative phase drift.

Sync gets the same treatment. Phone and Watch exchange versioned snapshots, acknowledge commands, retry when delivery fails and reject stale state. That is invisible in a screenshot, but it is the difference between a companion experience and two screens that merely look related.

Beto sat untouched for two months while work got busy. Then a user emailed about a real Watch bug, complimented the design and offered to help translate the app to Chinese. Around the same time, an App Store review asked for exact tempo instead of jumps of five.

That changed the framing. These were not hypothetical users; real people were depending on the tool. I carved out a Sunday and worked through the evidence.

App Store Connect reports 78% Day 1 retention and 33% Day 7 retention, both above its 75th-percentile benchmark. The cohort is still small, so I read these as encouraging signals rather than a conclusion.

The stronger outcome is behavioural: people cared enough to report a broken core promise, request finer control and offer a translation. That moved the roadmap from ideas I could imagine to problems the product had earned.

The strongest design decision was not the coral visual language. It was keeping the product small enough that its main task stayed obvious while the system expanded across devices, languages and more reliable audio behaviour.

I now model cross-device state earlier, test physical behaviours on real hardware sooner, instrument products from launch and treat shipping as the beginning of research rather than the end.

Building Beto also narrowed the gap between the product I can articulate and the product I can make real. That does not replace engineering partnership. It makes my specifications, trade-offs and conversations with engineers better grounded.