The best music I ever found, I wasn’t exactly looking for. Bought because the cover looked interesting and nothing else. I’d dig through a stack of stuff nobody wanted in the cutout bin and every so often pull out a treasure I’d never have found any other way.
In the late 00’s, I found bootleg blogs, and it was the cutout bin times a hundred. Someone would dig through estate sales and dusty record bins, rip the vinyl, compile the best tracks, and drop a zip file somewhere. Private press soul. Unreleased West African funk. Weird exotica from labels that pressed 500 copies and disappeared. I’d download it and fall into entire worlds I hadn’t known existed.
Then streaming became the easy way to find music. Apple Music and Spotify make their own bins for you to dig through, the home screen, the mixes, the weekly playlists, and it looks like the same thing. But their bins are pre-sorted. The wacky shit you used to stumble onto never makes it in, because it’s the same algorithm doing the sorting, and it already decided what you should hear.
The cutout bin never decided anything. It was chaos, a pile nobody managed, and the weird stuff survived in there precisely because nothing was curating it out.
But the music didn’t disappear. It’s all sitting right there in Apple Music, in Spotify, the cutout bin times a million now, catalogued and waiting. The pile only got bigger. The algorithm just had no reason to surface any of it. I kept finishing Radiohead tracks, so it kept serving me more Radiohead-adjacent sounds. Why would it randomly show me Orchestre Poly-Rythmo?
If you found your way to this, odds are you know the feeling. Your streaming app gets to know you so well that all it does is hand you more of what it already knows you like.
For me, that’s unacceptable. I’m the cutout bin guy. I don’t want the center of the thing I already love. I want the edges, and whatever’s past the edges. I want the chaos.
So I tried using AI to break out of it.
First, I installed an app called PlayTally, which allows you to see your Apple Music listening history across all your devices. I let it run a few months, then exported a CSV file for Claude.
- 4,631tracks analyzed
- 68%Rock / Alternative
- 702plays of “Tiny Dancer”
Dominated by 90s alt-rock and 2000s indie, and 0.2% of it from outside the US, UK, or Canada. The numbers told me everything I needed to know about the rut I was in.
When I tried to describe what I actually wanted instead, it wasn’t genres. It was textures.
- “Fender Rhodes dripping with chorus.”
- “Pedal steel in music that isn’t country.”
- “Quiet jazz, not the freeform stuff.”
“Alternative,” “Rock,” “Jazz” are huge buckets, and the micro-genres I wanted were drowning somewhere inside them. I knew they were in there. I needed a way to surface them from a system that wasn’t built to understand texture.
The thing I figured out, with a lot of help, is that the algorithm pays more attention to what you finish than to what you tap a heart on. So Claude built playlists around that. It paired an anchor I already loved with a challenge I didn’t know yet, close enough to follow. Radiohead’s “Pyramid Song” into Khruangbin’s “María También,” which shares the same Rhodes texture and slow build. Play them in order, let both finish, repeat. My read is that the repetition taught the algorithm to treat the two as related, and to start pulling more of the second kind toward me.
Plenty of it flopped. Japanese city pop went nowhere, Nordic jazz never took, and of the West African artists I seeded, only one stuck.
The cutout bin was a unique part of the analog era. Those boxes of chaos existed because some store owner dumped albums in a bin and sold them for a dollar to be rid of them. The mess was an accident. And there’s no incentive for such a bin to exist digitally. Which sucks, because I owe most of my musical taste to the chaos of the cutout bin.