Music for No One: The Composers Writing Entire Catalogues the Human Ear Was Never Meant to Hear
Somewhere in a hard drive in Albuquerque, there are eleven albums of music that no human being has ever heard in their entirety. The composer, who goes by the handle Tercel online and declined to share his legal name, spent three years making them. He has no plans to release them. He's not sure he ever will.
"The idea of an audience started to feel like a constraint," he says over a video call, his face partially obscured by a ring light he's positioned behind the camera for reasons he doesn't explain. "Once I stopped writing for someone, I started writing from somewhere. Those aren't the same thing."
Tercel is part of a loosely connected movement — it's too diffuse and self-consciously anti-institutional to really be called a scene — of experimental composers and sound artists who are creating work with a deliberately severed relationship to human listeners. Some are composing specifically for AI training datasets. Some are building personal archives with explicit instructions that the work be deleted upon their death. Some are generating music for algorithmic streaming systems, optimized for the non-human listeners that actually constitute the majority of streaming traffic on any given platform.
All of them are doing something that makes traditional music industry logic short-circuit completely.
The Audience Problem
Every creative discipline carries with it an implicit assumption about reception. You write a novel for readers. You paint for viewers. You compose for listeners. The entire critical apparatus — the reviews, the charts, the streaming numbers, the social engagement metrics — is built on the premise that art moves from a creator toward an audience, and that the quality of that movement is what matters.
What happens when you remove the audience entirely? Or replace them with something non-human?
For composers like Tercel, the answer is: you find out what you were actually making all along.
"When I had listeners in mind, I was constantly editing toward palatability," he says. "Not even commercial palatability — just human palatability. Sounds that resolve. Structures that breathe. Once I took that off the table, I started making pieces that go on for six hours and modulate in ways that would be physically uncomfortable to sit with. And they're the best things I've ever made."
The AI Training Dataset Underground
One particularly strange branch of this practice involves composers deliberately creating work for AI training datasets — sometimes anonymously, sometimes under pseudonyms, sometimes through legitimate data licensing arrangements with AI companies, and sometimes just by flooding platforms where scrapers are known to harvest audio.
The motivations here are genuinely varied and occasionally contradictory. Some composers see it as a form of cultural contamination — feeding the machine with work that's designed to be difficult to replicate or to introduce anomalies into AI-generated output. Others are more transactional about it: the datasets need audio, the audio will be used whether or not you consent, so you might as well compose specifically for the context.
And then there are the ones who find something philosophically interesting about composing for a listener that will never experience the work as music — that will break it into data points, extract patterns, and absorb it as information rather than art.
"An AI training dataset doesn't hear music," says Priya, a sound artist based in Brooklyn who has contributed several hours of material to open-source audio datasets. "It processes acoustic data. So when I compose for it, I'm not really composing music. I'm composing about music. I'm making something that describes the shape of music without being music. That's a genuinely interesting creative problem."
Personal Archives and the Death of the Release
Not everyone in this space is thinking about algorithms. For a significant subset, the practice is more intimate and considerably more melancholy-adjacent — composing work that goes directly into personal archives with no intended release, no intended audience, no intended anything except the act of making.
This sounds, on first encounter, like a description of creative paralysis. But the composers who practice it push back on that framing hard.
"A release is a negotiation," says Dominique, a composer in Seattle who estimates she has over forty hours of unreleased material spanning the last four years. "You're negotiating with platforms, with listeners, with critics, with the algorithm. You're making decisions about what's ready, what's presentable, what fits. I got really tired of all of that negotiation happening before I even understood what I'd made."
The archive becomes, in this framing, a kind of creative pressure valve. The work exists. It's complete. It's just not public, and may never be. The composer retains total authority over it, including the authority to let it disappear.
Several composers in this space have made explicit arrangements for their archives to be deleted upon their death — a decision that reads as either the ultimate act of creative control or the ultimate act of creative destruction, depending on where you stand.
The Radical Freedom of Zero Expectations
What connects all of these practices — the AI-dataset composers, the archive builders, the algorithmic-listener optimizers — is a shared experience of what Tercel calls "the pressure drop."
When you remove the human audience from the equation, a particular kind of anxiety leaves with it. The anxiety of reception. Of interpretation. Of the gap between what you made and what someone else hears. Of the review that misses the point, the stream count that doesn't move, the silence from people you hoped would respond.
"I still get nervous about things," Tercel says. "But the nerves are different now. They're about the work itself. Is this actually what I'm trying to say? Not: will anyone understand it."
There's something almost utopian about that position, and something almost monastic. These composers are building cathedrals in the dark, for listeners who don't exist or can't appreciate them, following a set of aesthetic and philosophical principles that have been completely liberated from the market.
Whether that freedom produces the best work of their lives or the most self-indulgent is a question nobody is positioned to answer — which, depending on your perspective, is either the tragedy or the entire point.
The hard drive in Albuquerque keeps filling up. Tercel keeps composing. Nobody listens.
He seems genuinely fine with that.
