01 / The Work
A time-aware music experience for the different states of the day. It adapts its scenes to the hour — movement in the morning, deep work at midday, unwinding toward evening.
Forge, 7–9am, is the scene opened most — before a run or a workout, for months now. Music for the body is not music for attention.
02 / Key Decisions
Five states, not a playlist.
Rejected a flat track list in favour of scenes mapped to the day's natural rhythm. The Work responds to when you open it, not just what you play.
Each scene has its own character: Begin is energizing, Deep Work and Flow ask for immersion, Unwind loosens into mostly jazz. With Forge at first light, each of the five has its hour.
One button, then silence.
The interface hides itself after a few seconds. Rejected persistent controls — the foreground should be the state, not the chrome.
Removing more than was added.
Track previews, switching animations, and hint text were deliberately removed across recent versions. Each removal was a judgment that the quieter version was the finished one.
03 / Model's Role
The music was generated with Mureka; the site's code was written by AI as well. The Work is itself the Lab's method in practice: the human role was not to write code or produce audio by hand, but to judge — which tracks fit each state, how structure and interaction should work, how scenes are edited and ordered, and when the result was finished. The model widened what could be built and what could sound right; the form was still a human decision.
Music created with Mureka as part of the work. Human direction, selection, editing, and product integration by Flow × Zen Lab. Availability and rights are described per release.
04 / Credits & Rights
- Music — Generated with Mureka. Rights depend on the account tier at the time of generation; see the rights page.
- Design & direction — Flow × Zen Lab. Code written by AI under human direction.
- Audio playback — Howler.js (MIT).
- Version — 1.1.2, released 2026.
05 / Limitations
- Music provenance — which tracks were generated under which prompts, what was rejected — is not yet published in detail.
- The experience is a web page, not a native app; background audio on mobile may be limited by the browser.