Listen to the Metal
A Kernel Noise Manifesto
Right now, somewhere, a Linux kernel is running. It is managing memory, scheduling processes, handling interrupts, negotiating between hardware and software at a speed no human can perceive. It is doing this on the device you are reading this on. It has been doing this since the moment you powered it on. You have never heard it.
Kernel Noise exists to change that.
What This Is
Kernel Noise is not data visualization with sound. It is not a monitoring tool. It is not generative music that uses system metrics as a random seed.
Kernel Noise is structural sonification — the act of making the internal life of the Linux kernel directly audible. Every sound you hear in a Kernel Noise piece is the direct consequence of a real system event. There is no random(). There is no sequencer. There is no artificial composition. The kernel composes. We listen.
If the system is idle, you hear stillness — a low drone, minimal and hypnotic, the sound of a machine at rest but never asleep. If load rises, the pitch shifts, the texture thickens. If the system panics — OOM killer, thermal throttle, kernel panic — the sound becomes what it should become: dissonant, industrial, honest.
The aesthetic is not “pretty.” The aesthetic is “true.”
If the Linux kernel is the invisible conductor of our digital world, then Kernel Noise is wiretapping it.
Why This Matters
We live inside systems we cannot see. The operating system is the most intimate technology in modern life — it mediates every interaction between you and your machine — yet it is designed to be invisible. We monitor it through dashboards, logs, graphs. We watch htop. We read dmesg. We stare at numbers.
I stopped watching. I started listening.
Sound is temporal in a way that vision is not. A graph shows you a snapshot. Sound shows you a process unfolding in real time. When you hear a system under load, you don’t just know it intellectually — you feel the weight of it. When interrupts cluster and scatter, you hear rhythm emerge from chaos. When the scheduler fights to maintain order against a flood of competing processes, you hear a conflict that has been invisible since the first operating system was written.
Kernel Noise makes this conflict audible. Not to make it useful. To make it real.
The Lineage
This idea did not appear from nowhere. It belongs to a lineage.
In 1952, John Cage walked into an anechoic chamber expecting silence and heard two sounds — his nervous system and his blood circulating. He concluded that silence does not exist. Every environment is full of sound; we simply choose not to listen. Cage spent the rest of his life building frameworks for paying attention to what was already there. 4’33” was not the absence of music. It was the presence of everything else.
Kernel Noise begins where Cage left off. The kernel is an environment. It is full of activity. We choose not to listen to it — not because it is silent, but because we lack the framework to hear it. Kernel Noise is that framework.
In the 1960s, Fluxus dissolved the boundary between art and life. George Brecht wrote event scores — simple instructions that reframed ordinary actions as performances. “Three Aqueous Events: ice, water, steam.” The Fluxus artists insisted that art was not a special category of experience but a way of paying attention to any experience. A dripping faucet could be a composition. A street crossing could be a performance.
Kernel Noise applies the Fluxus principle to computation. A context switch is an event. An interrupt is a percussion hit. A kernel compilation is a forty-minute composition that no one commissioned and no one intended. The art is not in what we add. It is in what we choose to hear.
In the 1990s and 2000s, Ryoji Ikeda began translating raw data — mathematical sequences, satellite signals, DNA code — into overwhelming audiovisual experiences. His work proved that data has an aesthetic dimension independent of its informational content. You do not need to understand the mathematics to feel datamatics. The signal itself carries beauty. Standing inside one of his installations in Paris last December — surrounded by walls of flickering data rendered as light and sound — I felt the direct connection between what he does with mathematical streams and what I wanted to do with the kernel. The difference is the data source.
Kernel Noise takes Ikeda’s insight and applies it to the most ubiquitous data stream on earth — the Linux kernel, running on billions of devices at this very moment, generating patterns that no one has ever heard.
The Architecture
The Truth of the Signal
There is no randomness in Kernel Noise. Every parameter — pitch, rhythm, texture, volume, density — is mapped from a real kernel event. CPU load, interrupt frequency, context switches, memory pressure, I/O wait, scheduler decisions. The mapping is the composition. The choice of which events to sonify and how to translate them into sound is where the art lives. Everything else is the system speaking for itself.
This is the first rule: if the system didn’t do it, you don’t hear it.
Post-Punk Bones
Kernel Noise borrows its structural philosophy from post-punk — specifically from artists like Pere Ubu, This Heat, and Public Image Ltd, who treated music as architecture rather than melody. In Kernel Noise, the kernel is the bass — heavy, omnipresent, foundational. The texture is the noise of computation — glitches, interrupts, the friction of hardware against software. The tension lives between the rigidity of the system clock and the chaos of asynchronous events.
Metal Box reimagined bass as a physical force. Kernel Noise reimagines the kernel the same way — not as an abstraction but as a presence you can feel through a speaker.
The System Is the Drummer
The Linux scheduler — the Completely Fair Scheduler — is an algorithm that tries to impose perfect order. Every process gets its fair share of CPU time. It is methodical, relentless, precise. It is Jeff Mills at the turntables — mechanical perfection, inhuman consistency.
But the peripherals do not obey the scheduler. Hardware interrupts arrive when they arrive — asynchronous, unpredictable, driven by the physical world. A network packet. A keyboard press. A DMA transfer completing. These are Sun Ra — cosmic chaos, improvisation against structure, events that refuse to be scheduled.
Kernel Noise is the sound of this conflict. We do not impose a time grid. There is no BPM. The rhythm is dictated by hardware interrupts. The system is the drummer. And like any great drummer, it plays differently every time.
The Rules
- No User Space Noise. We do not sonify the activity of any specific application. We sonify the impact of those applications on the kernel — CPU pressure, memory allocation, I/O contention. The kernel is the subject. Everything above it is context.
- Latency as Texture. If audio processing slows down the system, that slowdown becomes part of the sound. An XRUN is not a bug — it is the audible evidence of a system reaching its limits. Overload sounds like overload. We do not hide it. We feature it.
- Low Level First. We read from
/proc,/sys,tracepoints, andperfevents. We do not use high-level APIs that abstract away the reality of the hardware. The closer to the metal, the more honest the sound. - The Natural World Enters. The machine does not exist in a vacuum. Kernel Noise acknowledges that computation happens in physical space — in rooms with air conditioning hum, in data centers with fan noise, on a Raspberry Pi sitting on a desk near an open window. Field recordings — ocean, wind, birds, city noise — are the ambient layer that grounds the machine in the world. When the system is idle, you hear the world. When the system is active, the machine overtakes it. The boundary between natural and digital is the territory we explore.
What Comes Next
Kernel Noise v0.1 proved the concept — CPU load mapped to a sine wave on a Raspberry Pi with a custom DAC. It was raw, minimal, and honest. A machine speaking for the first time.
What follows is not an incremental upgrade. It is the full realization of the system described above — multiple kernel data sources layered into a unified sonic landscape, field recordings woven into the ambient layer, real-time generative audio that breathes between the digital and the organic. A five-to-ten minute piece that you can sit with, that changes as the system changes, that is never the same twice.
And eventually: a live performance. Kernel Noise on stage, running in real time, the audience hearing a system think. The invisible made audible. The noise made beautiful.
Pay Attention to the Noise
Every system has a voice. Every process leaves a trace. Every interrupt is a percussion hit that nobody intended. The Linux kernel has been composing music since 1991. Thirty-five years of continuous, unheard performance.
Kernel Noise is the microphone.
— Jihèd Ch, Paris, 2026
Kernel Noise is an ongoing project at kernelnoise.com. The source code, recordings, and technical documentation are open and available. If you want to listen to the machine, start here.