10,000 Songs, Onion Tone, AI Deviation
For our thirty-third issue, we recommend a consideration of Spotify’s new AI product features, a Yugoslav-American guitarist, and a revisiting of concerns within the text-sound music world and AI.
Recommendation: Yasunao Tone - “MP3 Deviation #7”
Spotify announced that it is planning to give users the ability to create AI-generated music. Co-chief executive Alex Norström said that the new AI tool would allow “one song to become 10,000 songs.” The feature will be available for an extra fee on top of a paid subscription. Spotify alerted the public at their 2026 Investor Day last week, recapped in a press release:
[SVP Charlie Hellman] announced landmark licensing agreements with Universal Music Group and Universal Music Publishing Group, enabling Spotify to launch a new tool where fans can create covers and remixes from participating artists’ and songwriters’ catalogs, with consent, credit, and compensation built in from the start.
The company also enumerated four “big ideas that will define [its] next chapter.” One of them is worth quoting:
The world operates as a power law. And for Spotify, that opens up significant monetization. As there is no “average user,” Spotify is building a portfolio of higher-ARPU (average revenue per user) products and add-ons to capture more value from our most engaged audiences.
So, in summary: The world operates as a power law, there is no average user, and one song can become 10,000 songs. Let’s take a moment to attend to the function of language here. It is provided with a mystical quality, partaking of a rhetorical style that has increasingly pervaded tech with the AI boom. Phrases like “feel the AGI” are thrown around as researchers from frontier labs tweet esoterically about the latest models. Even “vibe coding” suggests a sort of implicit mysticism. Sam Altman tweeted “build monuments in the desert” at the start of last year.
On to the “one song [can] become 10,000 songs” idea. I am curious about music’s precise status as a commodity in such a figuration. As we know, Spotify rents music to listeners. But I’m not used to being able to multiply things that I rent by 10,000, and digital music is not an apartment. Its material form as a commodity is a file that can be duplicated at very low cost. Its functional form is background experience-generator/sculptor. I don’t really know what it means to multiply that by 10,000 according to my preferences in a given moment.
If lots of people want to create lots of AI remixes of songs they love, then they relate to music in a historically new way, I think. Maybe this would make users both fans and artists. Maybe you are an artist if you use this tool, otherwise you would manage to generate many different types of music non-artistically, which would be an interesting achievement. Considering the larger context for such tooling, there seems to be traction around Suno, another AI music generation platform. Their CEO said in February that over 100 million people have used the app, and that they have more than 2 million paid subscribers.
But then again, it doesn’t take a musician to crank a spring-wound music box and make it play a song. Or trigger the “Going Down” announcement in an elevator by pressing a button. Obviously a lot depends on how the Spotify feature is built, but it is possible that an interesting prompt could lead to an interesting AI remix. The use of AI does not inherently make a song uninteresting. Maybe this is simply a new kind of musical hobby, or maybe it’s something else. If it ultimately proves to have nothing to do with art, then something strange is happening to art.
—Alexander Iadarola
Recommendation: Branko Mataja - “Tamo Daleko”
What can I tell you? I’ve just been sitting in my backyard most of the hours of most days, when I’m home. The weather is good, but this spring to summer period always feels uncanny. Something to do with the way time passes in the year and the way we recognize its familiar cadence, when excitement at the change of weather turns into almost a paranoid anticipation about its full expression in the long summer. I’m sitting out all the time now because I probably won’t want to do so much in a month. I’ll miss this month, I think.
This week, I recommend the Yugoslav-American guitarist Branko Mataja (courtesy of Sasha). It feels timeless in its slow articulation of slavic folk melodies pressed through the spacious, futuristic Les Paul-like tremolo/reverb. Mataja had an extremely difficult life. But he ended up in the suburbs of Los Angeles, where the weather feels a lot like it does in New York today.
I’m looking at my garden, where the onions and dill are going crazy. I look forward to the tomatoes and eggplants, but I guess the heart of the assortment are these perennial flavors that I don’t plant myself anymore. They just keep coming back with a stronger flavor, more sure of their place in the bed.
—Alec Sturgis
Recommendation: Yasunao Tone - AI Deviation #1 / Drake - Janice STFU
I’m in a Drake trench. This moment has been long-awaited since May 2024. The overall flow of my music intake has been blockaded by the three albums Drake released on May 15th, 2026. It’s complicated but also pretty simple. I’m trying not to pontificate on his music too much here. Not to blow up his spot, but Alex Iadarola and I were swapping texts about playing “Janice STFU” many times post-Iceman release. We also both happen to be writing about AI this week. Of course.
Instead, I’m revisiting the piece “Millenial Text-Sound,” I wrote a few weeks back. That writing addressed the various conditions that set up underground musicians’ fascination with mid-20th century experimental music—specifically Fluxus text scores and spoken word works—reimagined in digital contexts throughout the late 2010s and early 2020s.
There’s still a lot more work to be done here. The aim of expressing thetical goals and ideas through sampling, synthesis, text, field recording, and spoken word music composed within specific formal frameworks has clear parallels with the recent primacy of language prompting as the medium par-excellance of consumer-level AI. Text, as expressed through word scores, is its own generative prompting device for the composition of music. It’s basically a generative AI model, at least conceptually. “Algorithmic music” has been considered throughout music history: notation, scores, chance (Cage), serialism (Schoenberg), Bach & Hindustani classical music, scales and modes, automatic music, algorithms in musical software, computation, or just language following a model. With text specifically, works such as Jack Callahan’s Vocoder or Alvin Lucier’s I am sitting in a room demonstrate a 1-to-1 textual prompt entwined as a compositional process in the generation of music through the use of language.
Millennial text-sound music was an attempt at chronicling the cultural transition between our use of sampling, culture jamming, and collaging linguistic and musical memetic forms ahead of the arrival of full consumer AI. Sampling, synthesis, and arrangement in software were imaginative approximations of future computational and generative models—a kind of “dark ages” of pre-AI online image, text, and music making and circulation. Work was done to speculate upon a material and representational “context collapse” for music. It became almost an obsession in demarcating the stakes and appeals of music’s diminishing value (much discussed throughout the value collapse of physical media into streaming) dovetailing with independent music culture’s diminishing returns mired in exhaustion, trauma, and exposure. Text-sound music became a purely speculative space to playfully and seriously probe the collapse of music’s representational and material value—various captains going down with the ship, a kind of CPR for imbuing digital music with the breath of musical history.
Composer Yasunao Tone’s AI Deviations, which I had a hand in premiering at ISSUE Project Room in 2016, was one of the first attempts to connect Tone’s Fluxus background with AI technology, developed in collaboration with Tony Myatt, University of Surrey UK, and a team of researchers including Mark Fell and Dr. Paul Modler. His music is an open-ended translation process across media, technology, and writing systems, a noise media language that follows a progression of technological collapse experiments spanning the CD, the MP3, and more recently AI.
I think about this language, next to say, all the text wrought through Drake’s three recent albums (Iceman, Maid of Honor, Habibti). The Boy released pages upon pages of dubious, interesting, text—almost algorithmic in their adherence to scripts. Drake’s lyric on “Ran to Atlanta,” “I went St. John The Baptist on the Lam,” is a clever line that evokes the Saint who baptized Christ, the lamb of God. St. John the Baptist was also beheaded; and, Drake likens this to the “beheading” of a Lamborghini by riding it with the top off. Nice. Go ahead and ask an AI model to break that down for you. It replies: “Great question, Drake’s triple entendre suggests ...” with bullet points, em dashes, and a few ice cubes and owl emojis in there for good measure.
If we look at the last one hundred years of music, we’ll see a series of emergent avant-gardes and also a series of funerals. If we follow a thread from 20th century recording history, experimental music, digital music, independent music, and the advent of streaming and AI, it would look like a procession of 10,000 funerals, open caskets, wakes, and various forms of grief toward the total transformation of music and media as many have understood it. RIP Nina Protocol, but why look at this as a funeral?
Beyond Tone’s noise media language and St. John The Baptist Lamborghinis: What if I told you that music has only just begun?
—Nick James Scavo