Music as API: An Infrastructure Theory for AI-Native Platforms

A new class of music platform no longer sells recordings. It sells access — to generators, to catalogs, to moods — through interfaces that other software calls the way software calls software. The song has become a response to a request. This working paper maps what it means to take that literally: music as API, and listening as a logged transaction.

I. The Metaphor, Taken Literally

An API is a contract: specified requests, specified responses, uptime, metering, terms of service. The essay’s first section defines the metaphor with engineering precision and sets the discipline for everything after — every extension of the metaphor into aesthetics must survive contact with the literal thing, or be dropped.

II. The Stack Beneath the Song

The essay describes the AI-native music stack as infrastructure: a generation layer, a rights and provenance layer, a personalization layer, a delivery layer, and a listener-facing surface at the top. Each layer has its own owners and its own choke points. The song, on this account, is the last mile of a supply chain — and the listener is the only participant who cannot see the chain.

III. Who Owns the Endpoint

The political economy section, and the essay’s center of gravity. Infrastructure theory asks one question of every layer: who can deny access, and at what price. The essay applies the Institute’s capture-cycle framework — the pattern by which open protocols harden into tollgates — and asks where in the stack capture is already visible and where it remains, for now, only a design decision.

IV. The Listener as Client

An API logs every call. The essay reads personalization — the playlist that seems to know your afternoon — as the conversion of listening into telemetry: attention becomes request data, and request data trains the next response. What the interface learns is not what you love. It is what you ask for. The essay keeps that difference at the front of the argument.

V. What the Contract Cannot Specify

The essay’s defense of the residue. Some of what music does — grief at tempo, the room when a song lands, the request no listener would think to make — does not fit inside a request schema. The essay refuses both nostalgia and triumphalism. It asks which musical goods are genuinely uncallable, and which merely have not been called yet.

VI. Design Consequences

The essay closes constructively. If music is becoming infrastructure, the honest questions are infrastructural: provenance disclosure, portability, the right to an unlogged listen, and the terms on which human makers connect to the machine layers. The section sketches what an honest AI-native music platform would have to publish about itself — and what its silence would disclose instead.

Counter-case

The strongest deflationary view: API is only distribution’s newest costume. Sheet music, radio, the phonograph, the download, and the stream were each pronounced the death or the rebirth of music, and music outlasted every frame applied to it. On this view, calling music infrastructure mistakes a billing model for an ontology; listeners still hear songs, makers still make them, and the metaphor flatters platforms with a depth they do not have. The essay must show that something structural — not merely commercial — changes when the primary consumer of music is another machine, or the metaphor must be returned to marketing where it was found.

Sources to verify

• Jonathan Sterne, MP3: The Meaning of a Format — UNVERIFIED, confirm before citing in the finished essay
• Nick Srnicek, Platform Capitalism — UNVERIFIED, confirm before citing in the finished essay
• Tarleton Gillespie, Custodians of the Internet — UNVERIFIED, confirm before citing in the finished essay
• David Hesmondhalgh, work on the cultural industries and streaming — UNVERIFIED, confirm before citing in the finished essay
• Scholarship on music recommendation and algorithmic taste — UNVERIFIED, confirm before citing in the finished essay
• Industry and standards documentation for music-generation interfaces — UNVERIFIED, confirm before citing in the finished essay

Stakes

The Institute’s capture-cycle framework was built to describe how emancipatory infrastructures harden into extraction, and music is now its fastest-moving test case: a cultural form being rebuilt as plumbing in real time. This essay connects the Institute’s Film and Media Theory work to its Political Economy program, and it supplies the descriptive layer that the companion historical spine in this batch — the second pass on music-as-API — presupposes. Get the infrastructure description wrong, and every normative argument built on it inherits the error.