// the 10-second version
Your playlists are secretly a map of you. Taste Lab would draw that map, find the friend whose map looks most like yours, and point you to music you'd probably love but haven't found yet.
// the plan
What it would do
- Taste map: see which moods and genres you actually live in, not the ones you think you do.
- Music twin: compare with friends to find whose taste overlaps yours most.
- Taste drift: watch how your listening changes month to month.
- Edge picks: songs just outside your comfort zone: close enough to like, far enough to surprise you.
// how it would work
The pipeline I have in mind
- ConnectLog in with Spotify to read your top artists, top tracks and recent plays.
- DescribeTurn each artist into numbers from genres, popularity and listening patterns.
- MapSquash those numbers onto a 2D map where similar music sits close together.
- Compare and recommendMeasure overlap between people, and look for great music just past the edge of your map.
// the tricky part
Describing a song without hearing it
Spotify has restricted some of its audio-analysis data for new apps, so ready-made "energy" or "danceability" numbers are off the table. The challenge is building a good picture of taste from what's left: genres, artist connections and listening habits.
Still on the drawing board. Ideas, or want to be an early music-twin test subject? Say hi.