I started making music with Suno in November 2025. I thought I was making educational songs about scientific discoveries. I ended up making something else.

The first songs were straightforward — attempts to turn interesting science into memorable lyrics. But the models Suno was running at that time (4.5 through 5.5) were producing arrangements and sounds I had never heard before. A track I made for Jaws was the first moment I registered something unusual happening — the tool was reaching for sonic textures that felt less like generation and more like collaboration.

Then came The Tralfamadorian Trap. A Kurt Vonnegut-driven Memphis trap mashup, built in two volumes. That was the project's actual beginning. The combination of literary density and Southern hip-hop production unlocked something — the realization that AI music tools could be directed by careful conceptual framing, and that the output could carry real weight when the framing was serious enough.

From there I branched into other genres I love. Noir jazz. Industrial. French punk. Boom-bap. Neo-soul. Appalachian doom folk. The methodology emerged: identify a film, name an object inside that film that witnessed its central scenes, write a brief articulating what that witness would have registered, and let that brief direct the composition across whatever genre lens best fit the object's voice.

About six weeks before The Atlantic published its reporting on Suno's training data in June 2026, I started seeing artists on social media talk about their work being scraped. Melodies they had written appearing inside models they never authorized. The news that AI companies had taken what they wanted and sorted the legality later wasn't new — but hearing it in the voice of individual musicians was different.

One argument stayed with me. Melodies come from human beings. Not from computers. Every melody that ever trained an AI model was first worked out by a person connected to their own loss, love, joy, pain — some specific moment that made the shape of the sound inevitable. The models absorbed those melodies and strip-mined them of the context that produced them. I thought of an older artist I'd heard describe watching a younger artist steal his material. The scale is different. The structure is the same.

I thought about stopping. The question of whether these tracks were even copyrightable had already been settled — they aren't, not as pure AI outputs. Continuing meant doing it for reasons other than ownership. I kept going. The writing and the music were too good to walk away from. The results kept arriving in forms I had not asked for and could not have predicted, and crafting the prompts that produced them was its own creative labor.

I am an artist myself. I know what it costs to make something and send it into the world. The dilemma did not resolve. It just became part of what the project is.

The shift from "I am making music" to "I am documenting a methodology" happened gradually. The clearest moment came when I realized the relics could be revisited — the same object, the same witness-brief, rendered through different genre lenses as parallel readings of the same underlying material. That multiplicity made the project legible as conceptual work rather than catalog work. The tracks became evidence of a methodology, not products of one.

In March 2026 I filed Frame 25 with the United States Copyright Office. The filing claimed authorship through methodology — the creative tests, scoring frameworks, and editorial decisions that shaped the AI outputs under my direction. It was the first formal declaration that this was a project about process, not product.

In September 2026, Suno retired the models the methodology was built on. The newer versions produced results I could not use. The specific sonic territory the project had been exploring was no longer accessible through the tool that opened it. The project is now on hold with no new music being made.

What remains is the archive: roughly 500 tracks across more than 100 films, produced in 2026 within a narrow window of specific model availability. I am slowly moving into the visual side of this work — short films and animations built around selected relics, telling the object-witness stories the music began. A listening archive will follow.

Most of my current energy is directed into my own art practice: Earthlines, a drone photography practice revealing the figures and architectures embedded in geological formations. Frame 25 exists as evidence of a specific moment in AI music production and in my own working-through of what that moment required.