Woven All of Dream and Error

Author

Sasha Frere-Jones

Date

September 2025

About the text

Woven All of Dream and Error was initially published in S/FJ, Sasha Frere-Jones’ newsletter, in September 2025. It was a critical response to the album Woven All of Dream and Error by Kata Kovács and Tom O’Doherty.

About the author

Sasha Frere-Jones is a writer, critic, and musician, based in New York. He was the music critic for the New Yorker, and editor-at-large for the Los Angeles Times, and has written for numerous other publications.

Related works

Dream 1, Walk 1 (Anhalter Bahnhof)

‘Dream 1,’ one of the lathe-cut records in the series via which the album Woven All of Dream and Error was released

 

I’ve listened to Woven All of Dream and Error dozens of times since I first heard it in April of 2025—gleefully, of my own free will, as self-soothing. Plenty of music moves me and impresses me and ends up unplayed. This is different. Woven All of Dream and Error creates a tangibly human space and offers me a tactile blessings: sound that I do not get tired of being inside and around. In many ways, this music is a dialectical triumph because it is so sweet and mortal and yet suppresses many of its human traces. Tom O’Doherty’s guitar and bass parts hum and sustain and warp through various boxes, and Kata Kovács’s trumpet sails around inside digital echo, all of it lying between the train noises produced by machine learning. The tracks and wheels you hear are industrial vehicles that don’t exist but instead were generated by mulching the recordings of existing trains the artists never saw (open source recordings, YouTube, Internet Archive, etc). The technology of the locomotive was part of the 20th century’s military success story and AI—genocide’s new best friend—is central to the current military investment boom. Using the latter to recreate the former is a quiet way of melting the guns, so to speak.

Woven all of Dream and Error is a soundclash between the human and the machines that capitalism has nurtured under the banner of “convenience,” while using them most often for death (can’t get to the camps without rails) and surveillance (your phone is already using AI to log the fact that you’re reading this). As the duo told me, “the underlying tech can be applied to make cute pictures and also to murder families in Gaza.” Kovács and O’Doherty presented the original version of Woven at the Hošek Contemporary gallery, combining their audio with stills and movies of the railway lines around Berlin and Brandenburg they took over a five-year period. For some of the walks, Kovács and O’Doherty carried speakers amplifying the machine learning train sounds they developed with their colleague Kris Slyka. They then filmed that new walking performance. In person, the films and stills and sounds made for a sort of summoning that blends into memorial. As audio, though, Woven is what the duo call “a dub album in method.” They described their plan to “assemble a body of sounds, and shape them through subtraction, through iterative removal and recasting, through echo and layering.” They also described the “hallucinated trains” as “a new sort of dub versioning” that establishes a body of audio material, layer and subtract and ping-pong your original stems and then let the AI model learn from that modified seed. (Holly Herndon has called this process “spawning.”)

When the show was originally put on in September of 2024, the duo invited Dimitra Andritsou from Forensic Architecture to be a guest speaker for an event at the gallery. “We talked about the general cultural silencing in Germany and the wider lunatic turn in German public life,” they explained. “We talked about German weapon exports to Israel, which is Dimitra’s area of expertise, and how this connects to AI warfare technology like Lavender and similar IDF tools that have been developed in collaboration with German research institutions. We then talked about how the material we put in the exhibition—railways as remnants of war planning, past and future architectures of conflicts, and so on—connected, in turn, to all of that.”

“The world is woven all of dream and error” is the first line of Sonnet XXVI by Fernando Pessoa. Each track on the album is subtitled by one of the fourteen lines of the sonnet. The duo mentions that Pessoa’s invention of literary heteronyms like Alvaro Campos and Alberto Caeiro was also relevant, in his project of “inventing voices through which to speak, a kind of extended all-encompassing multiplicity of pseudonyms.” AI became, for Kovács and O’Doherty, a “machine for creating infinite masks and infinite stories,” as Pessoa’s trunk full of scraps had been for him. This does not mean, however, that their use of AI is about creating some sort of vague, reassuring dreamscape (even if this album is legitimately pleasurable). “We had the intention with the album of foregrounding the technology and explicitly linking one process to its technological antecedent: AI/ML networks and, respectively, train networks,” the duo said. “This artistic decision is rooted in a materialist way of looking at the world, in the Marxist sense of materialism.”

Kovács and O’Doherty live and work in Germany, the facts of which influence the nature of the work, as it was generated, and how they discussed it during the original exhibition. This is a different but ineluctable aspect of materialism—place and how it shapes the political economy of the art produced there. That also affects how the AI model they built with Slyka works. “Dreamsloth,” as they call it, processes sounds in an idiosyncratic way. “One thing about the hallucinated train noises is that they would sometimes get stuck doing a train-going-into-tunnel-style shoooommm noise,” O’Doherty told me. “There’s no logical consistency to the sounds, they’re just the result of computational guesswork—there would never be a corresponding whooooosh of a train coming out of a tunnel. The program ended up just repeating the first part, like a contemporary variation of a locked run-out groove or skipping CD.” The duo also pointed out that the program mostly spit out “quite midrange-y sounds” and they had to “do some persuading of Dreamsloth to give us any bass.”

But they did, and it’s a truly magical thing they’ve made. The train audio is something like an ocean, or a buffering wave, not at all like the pink noise bricks you get in much ambient music. There’s a lot of clean space between the sounds—it’s not a murky or a lazy arrangement of events. Kovács’s trumpet lines are long, confident presences, like canopies being stretched over the sand. The bass and guitar parts could easily be organs or vibraphones, extended and amplified. They form small chords that waver and grow and disappear. The train chatter doesn’t always sound like a train, though when it does, it seems a bit like our chords and tones are static and the train simply passes through what we are doing together. O’Doherty said the rails and track sounds “remind me of the feeling of being on the train to school in the mornings when I was a teenager, when I would be listening to music and trying to will the sounds of the rails to line up with the beat of whatever I was listening to.” Woven All of Dream and Error has a way of lining me up, again and again.