This is incredible. Now that you've done it, it seems obvious. But I hadn't thought of it before since we have so many methods of music discovery these days already. Despite this glut of competing discovery mechanisms, I've been able to find a playlist full of new beautiful music in a genre I've found really hard to articulate to "hard" systems in the past or fully understand myself. This "soft" intelligence is fantastic.

If you can do the reverse as well... to recognize abstract themes in a playlist... you have a huge thing here. Those could be taken and combined or adjusted to perform automated playlist curations in an amazing way. Spotify should be hiring you right now. Imagine being able to pick an existing mood playlist you love and say "more with this feel... but higher energy."... being able to pick "rainy day" playlist and curate between "hopeful only" or "meloncholoy only".

Or being able to pick the top two thematic components of two people's playlists and not just interleave the songs but generate a new list with a shared mood. Electronica friend and Hip Hop friend hanging out? No more "one electronica, one hip hop" interleave on merged playlists... all songs with both string electronica and hip hop elements.