Tech serves art

We use AI to make the platform smaller, not the artist.

Walk through most of what's being sold as "AI for music" and you find the same trick underneath. Generated tracks. Invented artists with invented life stories, padding out playlists. The point is always to cut the human out and keep their cut of the money. Cassette will never sell AI-made music, and never dress a machine up as a person who made something. There's no version of that line that bends.

But the easy take, that AI is bad, full stop, misses the part that changes who's in charge.

For most of the internet's life, size was the wall. Building anything that could go toe to toe with Spotify took hundreds of engineers and a pile of money, which meant a few companies could do it and everyone else couldn't, which meant those few set the terms for the artists and for you. That was the entire source of the middleman's grip. You needed them because the thing was too expensive to build any other way.

That wall just fell. A small team can now build software that used to demand an army, for a fraction of what it once cost. And once a small team can build the thing, the giant has lost the one advantage that ever mattered. That power comes loose, and it can slide back toward the people who make the music and the people who pay for it, instead of pooling up in whoever was big enough to stand in the middle. That's the bet.

So AI gets used here the way you'd use any sharp tool, to do work that would otherwise need a budget that doesn't exist. It reads through millions of artists and albums and drafts a first description. It works out which artists genuinely sound alike, and why. It keeps a catalog this size searchable without an army of catalogers. Every artist it touches is a real person who made real music. The machine describes and connects human work. It never stands in for it.

One version of this technology jams the power even tighter into a few hands. Another version cracks it open. The second one is the one worth building.