The Challenge
YouTube wants your attention. Autoplay, recommendations, comments, suggested videos. The platform fights to keep you watching.
But sometimes you just want to listen. Background music for deep work. A three-hour mix while you write. Focus audio without the focus-stealing interface.
The Approach
What if a music player worked like vintage radio equipment? No endless feeds. No recommendations of things you've never heard. Just your frequencies, ranked by how you actually engage with them.
The interface itself becomes invisible. All hardware switches and amber glow, nothing begging for your attention.
The Solution
A distraction-free YouTube audio player that looks like Cold War radio equipment. Save videos as "frequencies." Mark interesting moments as "signal peaks" and jump between them with a keystroke.
The app tracks your actual listening behavior. Not stars you'll forget, but how many hours you've spent with each piece. An optional AI "archivist" writes museum-placard descriptions of your collection, so you can experience music in a new way.
The Outcome
Background listening without the background noise. You control what plays. A TikTok-style ranking algorithm surfaces what you actually gravitate toward based on stars, skips, and completions.
Your position saves automatically. Your data stays on your machine. The interface disappears, and the music stays.
What I Learned
Behavior reveals preference. Asking users to rate things creates meaningless data. Watching what they actually do, what they return to, how long they stay, tells the real story.
You don't need ML. Research from ByteDance and implicit feedback papers showed that simple weighted averages of real behavior outperform complex systems. Stars, skips, and completion rates are enough.
Constraint creates personality. One accent color. One typeface. Terse, uppercase microcopy. These limits forced every design choice to earn its place, and the result feels intentional rather than decorated.