Independent music where art
and technology meet.
About Wired Moth
Wired Moth is dedicated to the strongest new music being made with code, custom software, machine learning, live systems and other unusual technologies. We cover pop, club music, composition, sound art and everything in between.
The music always comes first. We are interested in technology when an artist uses it to make new creative choices: building an instrument, writing a set of rules, training a model on their own material, performing code or turning data into musical form.
The site is updated regularly. Fresh Signals gathers the newest releases, while the 2026 archive reaches back to January and will feed our end-of-year selection.
We welcome any feedback regarding the site and the music we are showcasing. We encourage you to support the artists and labels who are releasing the music by purchasing music on Qobuz and Bandcamp or streaming where available.
Latest feature · 9 October 2026
Omar Medina
The Mexican composer’s electronic systems leave room for uncertainty. His story gives that uncertainty a shape.
Migration, improvisation and sonic storytelling. Inside Medina’s approach to generative systems—and the personal history behind GOLEB.
Previous lead signal · September 2026
DJ_Dave
The work stays visible.
A three-minute club track made by writing and performing code in Strudel. Its directness says more about DJ_Dave’s practice than a technical demonstration ever could.
DJ_Dave has built her live identity around disclosure: the audience sees the code as she writes it. A mistyped character can change the music; a working line can be pushed, rearranged or abandoned. The screen shows the labour, taste and risk that a finished electronic track usually hides.
“Await” was released by Mad Decent on 4 September. DJ_Dave identifies it as coded in Strudel, while interviews describe how she uses Strudel and Sonic Pi to control samples and synthesis, including vocal chops and arpeggios.
The track is structured as pop rather than a software demonstration. A recurring vocal fragment supplies the hook, the tension builds quickly and the song ends before the idea loses force. The code is central to how it moves, but the judgement is a songwriter’s.
Newly released music that has earned its place here. The window covers 90 days; older selections remain in the 2026 archive.
Take a closer listen. Choose Listen here to play an official preview or open a release’s Bandcamp or YouTube player without leaving Wired Moth.
01≤ 90 daysA chamber opera built from memory behaving badly. Trumpet, voices and ensemble writing are sampled, fractured and folded back into themselves until acoustic detail and electronic afterimage become hard to separate.
Reid and her collaborators developed the work across five years using interactive electronics in Max/MSP. Recordings were repeatedly broken apart, sampled, reshaped and returned to the composition, with Reid directing the iterative form and Ryan Gaston credited for programming. No machine-learning claim is made.
Process source
02≤ 90 daysData activism becomes vivid electronic composition rather than a dashboard with a soundtrack. Climate, inequality and social metrics move through dense synthetic colour, while one human producer keeps the album's argument, pacing and emotional weight in view.
Portrait XO sonifies data from Our World in Data tied to the UN's Sustainable Development Goals, then combines it with sounds made through consent-based models of her own work and material from WOULG and Rob Clouth via Neutone AI and Datamind Audio. She composed, recorded, produced and mixed the album; the models are credited instruments inside a human-authored work.
Process source
03≤ 90 daysA sixty-three-minute sequel whose dark ambient mass keeps revealing moving parts: coded percussion, granular guitar and dense electronic weather arranged with the patience of long-form composition rather than the passivity of a system left running.
Roddy uses TidalCycles compositionally, controls granular and effects chains from guitar, and deploys generative systems he designed in Max/MSP. He wrote and performed the full work, retaining authorship through the systems' construction, instrumental control and arrangement. No machine-learning claim is made.
Process source
04≤ 90 daysMoonlight becomes control voltage, but the appeal is not the diagram. Across a single slow arc, cracked electronics and luminous drones keep changing density as two performers steer and answer a signal that can never quite repeat.
Simon steers homemade telescopes fitted with photodiodes and optical sensors. An analog amplifier turns their changing readings of celestial light into control voltage for Geologist's modular synthesizer, modulating pitch, filters and other parameters while Geologist responds in real time. The performers built and direct the system; no machine-learning claim is made.
Process source
05≤ 90 daysA restored vocal Eno album where songs, treated voices and ambient detail keep exchanging foreground and atmosphere. The generative element is track-specific rather than a claim for the whole record, but its soft instability matters inside one of the album's most suspended moments.
On Going Unconscious, Eno used SSEYO Koan to generate a textural pad, placing a software-authored layer inside a deliberately composed and produced song. Elsewhere he sang, wrote and shaped voices from a keyboard through harmony, formant and pitch controls. This is a 2026 reissue of a 2005 album, not newly composed 2026 music, and no machine-learning claim is made.
Process source
06≤ 90 daysFive small songs suspend soft voices and electric bass inside a haze of clipped code, guitar reverb and damaged air. The slowcore intimacy survives every digital fracture; even the glitches seem to breathe with the singers.
Seina's live coding is paired with Kleo's electric bass and the duo's vocals across all five pieces. The coded electronics generate and reshape the ambient structures and glitch processing around plainly human performances and songwriting; no machine-learning claim is made.
Process source
07≤ 90 daysFive concise pieces move between glitch, IDM, drone and electroacoustic pressure without sounding like demonstrations. Each system sets a situation in motion; Medina's improvising ear supplies the friction, pacing and sense of arrival.
Medina developed a separate generative system in Max/MSP for every composition, then shaped its behaviour through improvisation with modular synthesis, field recordings, processed samples and electronic instruments. He wrote and produced the finished album; no machine-learning claim is made.
Process source
08≤ 90 daysThe premise is funny and exact: Steve Reich’s Four Organs crossed with Louie Louie. A thirteen-piece band turns Orcutt’s software-generated structure into loud, unstable ensemble music, where strict timing keeps threatening to come apart.
Orcutt designed the open-source Cracked software that generated the structure of The Four Louies by combining material from Steve Reich's Four Organs with Louie Louie. This live recording realizes that computer-built composition through a thirteen-piece ensemble recorded at Roulette in Brooklyn; no machine-learning claim is made.
Process source
09≤ 90 daysTwo long sides of slow-motion instability. Tones gather, thin and drift without settling into a loop, while tiny changes in pressure make the custom instrument feel less like a machine demonstrating chaos than a landscape being patiently played.
Heel combines a self-built Eurorack instrument with custom software whose modulation is derived from chaotic attractors. The system produces structured, continuously varying behaviour without loops; Heel shapes its output through composition, intuition and performance. No machine-learning claim is made.
Process source
10≤ 90 daysHundreds of tape fragments become an album with warmth, swing and abrasion rather than nostalgia. Chance throws rhythmic and harmonic material together; Brett Bullion's ear turns those collisions into eight pieces that feel loose-limbed but decisively cut.
Bullion recorded and hand-spliced hundreds of analogue tape loops containing rhythmic, harmonic and textural fragments, then used a chance procedure to select combinations. He transformed the results through speed changes, filtering and studio arrangement, making the random-selection rule materially formative while retaining human authorship. No machine-learning claim is made.
Process source
11≤ 90 daysCello and synthesis meet as materials rather than opposites. A single long piece moves through bright partials, unstable beating and hard-edged electronic mass, turning spectral analysis into music with real physical tension.
Railton's cello spectra were analysed and digitally resynthesized as collections of partials. Eilbacher and Railton calculated the distances between those frequencies to construct the harmonic system, then arranged simplified synthetic tones against the performed cello. The artists designed the analysis, synthesis and form; no machine-learning claim is made.
Process source
12≤ 90 daysEarly trance and a Catholic pipe organ turn out to share more than repetition. Denuc uses a custom sequencer to make the organ move at inhuman speed, then lets the church’s acoustics give those patterns weight and duration.
Denuc composed and programmed the album for a MIDI-controlled pipe organ. His hand-built Max/MSP sequencer and OpenSoundControl system address the instrument's keyboards, stops and pedals, producing rapid repetitions and other passages beyond human performance while Denuc shapes, records and edits the result. No machine-learning claim is made.
Process source
13≤ 90 daysForty minutes of dance music that keeps changing the floor beneath itself. Clashing tempi, snapped digital percussion and strange accelerations make Fell's formal systems feel less like abstraction than a new physics for bodies.
Boomkat documents poly-temporal permutations, asymmetrical rhythmelodic arrangements and Fell's long-held ideas of multi-stability. Patterns appear to adapt against one another in real time, while acceleration, deceleration, attack and decay become compositional variables. Fell authors and controls the system; no machine-learning claim is made.
Process source
14≤ 90 daysThree hours of pastoral drift generated by an unusually poetic control system. Text becomes Morse-code gates, lunar phases disturb their timing, and three resonator chords open into shortwave haze: ambient music whose apparent stillness is full of machinery.
Brian Grainger recorded the piece on his portable Audioholistics modular system. A module translates a section of Homer's Odyssey into twice-daily Morse-code gate patterns; a second module subdivides those gates differently each day according to lunar phases. Shortwave radio passes through feed-forward delay and distance-based filtering. Grainger chose and patched the system and recorded its evolving output; no machine-learning claim is made.
Process source15≤ 90 daysA three-minute club track with a vocal hook, quick-building tension and no wasted motion. The code is audible in its clipped repetitions and abrupt turns, but the decisions are those of a sharp pop songwriter.
DJ_Dave's first-party performance post identifies Await as music coded in Strudel. Her August SoundCloud interview describes Strudel and Sonic Pi as instruments for triggering samples and synthesis parameters, including randomized vocal chops and arpeggios. No machine-learning claim is made.
Process source
16≤ 90 daysTen electric guitars behave like a flock with faulty instincts: copying, mutating and recombining until the score becomes a weather system. Nearly two decades in the vault have not cooled its feral group intelligence.
Sharp's 144 composed cores are distributed across twelve score sheets, then activated through rules for imitation, addition, recombination, transposition and mutation. The system models flocking birds, cellular automata, hunting packs and recombinant amino acids; this 2026 release documents a 2008 performance. No machine-learning claim is made.
Process source
17≤ 90 daysEight pieces of intricate software music where glassy melody, broken rhythm and microscopic debris keep swapping foreground. The code never arrives wearing a lab coat; it behaves like an instrument with private muscle memory.
EVEL ties each piece to processes developed inside Michael Cella's own creative environment. Artist-authored compositional systems organise structure, texture and evolution, with software design and musical creation treated as one practice. No machine-learning claim is made.
Process source
18≤ 90 daysA lean, two-part study in sequence and surface. Twelve minutes is enough for clipped pulses, sustained tones and rough-edged digital movement to keep changing their relationship without losing the thread.
Both tracks were recorded in real time while v4w.enko configured a generative sequencer programmed in Max/MSP 4.5.7. The artist builds procedural instruments in Max/MSP and Jitter, placing the authorship in the design and live handling of the system. No machine-learning claim is made.
Process source
19≤ 90 daysAcoustic performances are taken apart without losing their grain. Soft drums, guitar, brass and mallets keep surfacing inside slow-moving polyrhythms: patient headphone music with real warmth and rhythmic life.
Every source sound is an acoustic performance recorded by the duo or their collaborators. The label documents custom-built software processing, while Adam Parkinson describes a Max/MSP patch he jams with as a musical instrument; Jordan Cohen then cuts and resamples the material into paired pieces and shifting rhythmic patterns. No machine-learning claim is made.
Process source
20≤ 90 daysA ten-track rhythm study that keeps changing the grid without losing the body. Polymeters slide across one another, subdivisions breathe and trance patterns mutate in place: intricate electronic music with forward motion rather than technical display.
Swaroop Guhathakurta translates layered, interlocking drumming principles into generative electronic structures built from shifting time signatures, polymetric and polyrhythmic relationships, evolving subdivisions and interdependent patterns. Those structures create continuous variation instead of fixed repetition; Guhathakurta wrote, produced, composed and mastered every track. No machine-learning claim is made.
Process source
21≤ 90 daysTwo long real-time documents with the casing left open. One catches Whitman’s computer music in public motion; the other lets hybrid digital-analogue patterns spread across the room after hours. Process music as lived-in voltage, not white-coat proof.
The release joins a live performance identified by Whitman as Computer Music with a later set of hybrid digital-analogue synthesizer patterns. His official site describes the surrounding practice as improvised live electronics captured in real time and research into generative and process systems across analogue and digital domains; no machine-learning claim is made.
Process source
22≤ 90 daysThree compact transmissions from a laptop rabbit hole: clipped percussion, synthetic pressure and sudden negative space assembled while the code is still warm. It behaves like techno that has learned to edit its own nervous system in public.
Stotrojka identifies _semism_ as Semi Ismaili's live-coding project and ties this debut EP to software, sound and real-time processes used to build continually shifting electronic environments. All music is produced by Ismaili; no machine-learning claim is made.
Process source
23≤ 90 daysTwo walks become a duet across distance. Field recordings and recollected voices give Law's turntable performance an intimate, restless form: places are revisited, interrupted and brought into contact, with the accompanying walks opening the source material to a second hearing.
Law's Territory interface maps recorded walks, photographs and oral recollection to the position of a turntable needle. Needle movement, direction and vinyl manipulation reshape the walks in real time. This release documents his March 2026 Algorithmic Art Assembly talk and performance, alongside the San Francisco and Glasgow walks. Law authored and performs the system; no machine-learning claim is made.
Process source
24≤ 90 daysSix compact pieces move between ambient drift, clipped rhythm and indie-electronic melody without treating the code as a spectacle. Gold Kim’s programming feels closer to touch than demonstration: structures change in the moment, but the record keeps a warm, handmade scale.
The release was composed and recorded through live-coding performance, with Gold Kim writing and modifying code in real time. Kim programmed the drums, bass, synthesisers, keyboards and effects, then handled composition, sound design and the album's direction. The code is the instrument and Kim remains the author and performer; no machine-learning claim is made.
Process source
25≤ 90 daysFourteen research careers recast as bright, patiently unfolding electronics. Citation counts become pulse and pitch; Karhide then trims the spreadsheet chill with grain, drift and enough melodic sentiment to keep the graphs breathing.
Sonimet maps each researcher's publication timeline, career milestones and citation density into MIDI velocity, pitch and rhythm. Karhide imports those generated parts into Ableton Live, then arranges and reshapes them with Synplant, Microtonic and effects. The release documents data sonification and substantial human production, but makes no machine-learning claim.
Process source
26≤ 90 daysThirty-three pieces form a sprawling atlas of computer composition: brittle pulses, skewed counterpoint, low drones and quick synthetic studies. Its scale is unruly, but the changes of method keep it closer to a sketchbook than a software dump.
Quevedo Gómez links the pieces to SuperCollider and SCAMP-Python systems using stochastic processes, chaotic models and algebraic or mathematical procedures. He authored the programs and compositions across the collection; no machine-learning claim is made.
Process source
27≤ 90 daysFour ten-minute drones from one Bitwig Grid organism: less ambient wallpaper than a controlled weather front. The edits slow, darken and roughen the same orbit until repetition starts behaving like depth.
Polarity built the release from a single generative Bitwig Grid patch. The official release notes identify four versions of that patch and link both the original performance and the downloadable preset, making the system behind the music inspectable rather than merely claimed.
Process source
28≤ 90 daysEight long polyrhythmic chambers built from raw synthesis, harmonic pressure and the kind of digital repetition that keeps finding trapdoors. Severe enough for the lab, physical enough for the room.
The release page describes generative, procedural sound lines, constant rhythms and polyrhythmic structures stitched into finished compositions. v4w.enko identifies the underlying practice as real-time algorithmic music made with self-programmed Max/MSP/Jitter instruments; no machine-learning claim is made.
Process source
29≤ 90 daysEight electroacoustic problems with no interest in resolving politely. Noise, concrète debris and unstable pulse keep changing the terms of engagement; the algorithms supply pressure, not personality.
Najafi builds each piece around a different algorithmic or generative system, moving between computer music and electroacoustic experiment. EVEL ties those systems directly to the album rather than merely describing the artist's wider practice.
Process source
30≤ 90 daysSeven compact shocks written in the browser, where rhythm behaves like syntax under pressure. This is live coding with the screen glare left in: brusque, unstable and far too physical to feel like a demonstration.
c_robo_ works in Strudel, the browser-based implementation of TidalCycles. The label ties this release specifically to live coding and algorithmic composition: patterns are written and transformed as code, while the artist's timing and structural decisions shape the finished record.
Process source
31≤ 90 daysA compact TidalCycles session that values momentum over exposition. The main take and two earlier passes keep the same material unsettled, letting clipped rhythm, repetition and small mutations do the structural work.
th4 identifies Strand as a TidalCycles improvisation recorded in August 2026, with two earlier takes included as bonus tracks. The code is used as a live performance environment for shaping patterns and variation; no machine-learning claim is made.
Process source
32≤ 90 daysSol LeWitt’s grid is shoved through a speedcore lathe until concept, kick drum and counting error become indistinguishable. Four blunt proofs that informational music can still leave bruises.
Code translates the lines and spaces in Sol LeWitt’s 1970 drawing Six Thousand Two Hundred And Fifty-Five Lines into 6,255 speedcore kicks, rendered at 33 and 45 RPM; two versions add 6,255 Hz drones. The rule is the score, and the release is its deliberately excessive execution.
Process source
33≤ 90 daysFour slow forests grown while nobody was watching. Analogue tones coil, flower and lose their edges across seventy-seven minutes; less a set of ambient tracks than a weather system caught deciding what shape to take.
Riedl assembled the album from days-long generative sessions recorded between October 2025 and January 2026. He set analogue synthesis and drifting effects in motion, left the machines running unattended, then returned periodically to add textures or adjust settings; the released pieces are edited excerpts from those evolving systems.
Process source
34≤ 90 daysFive minutes of data made physical. Ventimiglia treats digital structures as raw sonic matter, listening for the configurations the system produces and choosing where observation becomes composition.
Ventimiglia converts digital data directly into audio without a conventional sound source. Different data structures produce a catalogue of system outputs; the artist observes and selects the resulting sonic configurations rather than composing noise conventionally. No machine-learning claim is made.
Process source
35≤ 90 daysA box of VCV Rack mechanisms taught to twitch, disperse and occasionally eat their own wiring. Tiny rhythmic particles become a loose, quick-witted electronic ecosystem rather than a preset parade.
Each piece uses a different VCV Rack patch built mainly from Audible Instruments ports of Mutable Instruments modules. Stochastic trigger trains, split modulation streams, feedback networks, cross-patched sequencers and real-time muting turn each system into an instrument for improvisation.
Process source38 selections from earlier in 2026, plus 3 recent records kept for context.
Live coding
Mad Decent · 24 Jul 2026
Neural audio
EVEL · 24 Jul 2026
Generative software
Constructive Music · 24 Jul 2026
Live coding
Self-released · 21 Jul 2026
Generative software
EAT DIS · 18 Jul 2026
Generative software
Self-released · 17 Jul 2026
Algorithmic composition
Nearest Neighbors · 9 Jul 2026
Live coding
Self-released · 29 Jun 2026
Algorithmic composition
EVEL · 19 Jun 2026
Neural audio
Self-released · 10 Jun 2026
Rule-based composition
Redshift Records · 5 Jun 2026
Live coding
Self-released · 31 May 2026
Algorithmic composition
Self-released · 29 May 2026
Algorithmic composition
EVEL · 15 May 2026
Live coding
Mad Decent · 8 May 2026
Rule-based composition
Elli Records · 28 Apr 2026
Live coding
Self-released · 16 Apr 2026
Live coding
Self-released · 15 Apr 2026
Generative software
EVEL · 10 Apr 2026
Algorithmic composition
Self-released · 6 Apr 2026
Generative hardware
Self-released · 16 Mar 2026
Generative software
Fönstret · 15 Mar 2026
Algorithmic composition
Cloud Recordings · 14 Mar 2026
Generative software
Self-released · 6 Mar 2026
Generative software
Self-released · 6 Mar 2026
Generative software
Old Technology · 6 Mar 2026
Live coding
Mad Decent · 27 Feb 2026
Algorithmic composition
Highpoint Lowlife · 20 Feb 2026
Neural audio
Atlantic Records Spain · 20 Feb 2026
Live coding
Self-released · 19 Feb 2026
Live coding
Self-released · 26 Jan 2026
Algorithmic composition
Self-released · 23 Jan 2026
Live coding
Self-released · 20 Jan 2026
Rule-based composition
Self-released · 15 Jan 2026
Live coding
Bio Future Laboratory · 14 Jan 2026
Live coding
Self-released · 10 Jan 2026
Neural audio
577 Records · 5 Jan 2026
Algorithmic composition
時の崖_tokinogake · 4 Jan 2026
Machine listening
Classical Computer Music · 12 Dec 2025
Machine learning
Outside Time · 19 Sep 2025
Algorithmic composition
Brainfeeder · 30 May 2025
Three records, three different relationships between an artist and a system. These are entry points, not rankings.
Arnal built AMA from vocal improvisations, adding words later, and developed models of her own voice with the Barcelona Supercomputing Center. The technology lets one singer become a choir of related but unfamiliar selves. What gives the record its force is the human material around that system: breath, memory, grief and Arnal’s decisions about which synthetic voices belong beside her physical one.
Early trance and a Catholic pipe organ turn out to share more than repetition. Denuc’s hand-built Max/MSP sequencer controls the organ’s keyboards, stops and pedals, producing passages no organist could physically play. After a year working with the instrument at Sankt Antonius in Düsseldorf, he made music in which rapid patterns gain weight from air, wood and the long acoustics of the church.
Orcutt’s open-source Cracked software combined the structure of Steve Reich’s Four Organs with material from “Louie Louie”. For this recording, the computer-built piece is transcribed for thirteen players. The joke survives, but so does the friction: a rigid framework becomes loud, unstable ensemble music once human timing and touch enter the room.
Performances, installations and research works live here. They matter—but they are not passed off as commercially released tracks.
Research performance · 3 Oct 2026
A music transformer trained on Rudess's own performances listens and answers him in real time, while generating material for the MIT Chamber Chorus. The machine predicts; Rudess steers, restarts and decides.
View projectResearch demonstration · Mar 2026
Eight EEG bands steer synthesis, spatialisation and algorithmic transformations inside a pre-composed spectral frame: thought as control voltage, not mysticism.
View projectLive performance
Neural synthesis pushed into a fluid stage work, developed through IRCAM's Forum programme.
View projectVoice model
An artist-trained RAVE model fed by Blonk's extended vocal vocabulary: splutter, scrape and impossible throat geometry.
View projectInstallation
A motorised music box, electroacoustic composition and neural audio processing occupy the same fragile object.
View projectNeural sound work
RAVE processing reframes everyday listening through a study of mediated attention and perception.
View projectWired Moth is interested in what happens when musicians give technology a meaningful part in the creative process — and something human comes back out.
Code can write rules. Software can generate possibilities. Models can learn. Systems can behave unpredictably. But the records we care about still carry the fingerprints of the people making them: judgement, obsession, humour, mistakes, emotion and taste.
Sometimes the machine is almost another performer. Sometimes it is an instrument the artist has built. Sometimes it is simply a system capable of taking the music somewhere its creator couldn’t have reached alone.
Not everything here is warm or obviously human. Some of it is deliberately synthetic, mechanical or strange. We will make room for that too when the idea or the artist is compelling. But technology alone isn’t enough.
The music has to have a soul.
Wired Moth works the same way. Machines help us search, research, verify and discover. Human ears make the final decision.
Music first. Human judgement always.