Fernando G. Tamajon — the sound artist working under the names AudioWanderer and MALAVENTURA in southern Spain — spent twenty years building experimental generative music tools: Lemur templates, Lua scripts, Arduino programs, all freely available online. But he recently made something completely different: the AW Noise Maker, a palm-sized handmade device in a self-made box. Inside the box is a handwritten slip of instructions — and a warning. This is "a non-idiomatic synth made not to be controlled or understood in any manner."

The device has four controls: two buttons, two knobs. One button shuffles the generated tones; another freezes the current sound into a short loop. The knobs handle filter and LFO speed — but their effects aren't stable, because the random FX algorithms reinitialize on every boot. The OLED screen displays glitch patterns that don't correspond to any parameter state. This isn't accidental unpredictability: it is deliberately engineered incomprehensibility.

In SAE terms, the electronic synthesizer is an already-construct: its implicit promise is learnability. Even the most chaotic-seeming instruments — the Rungler, the Lyra-8, the Strega — reward invested attention. Time in, control gained. Within the chisel-construct cycle, "controllable chaos" is already a complete category with a name, a community, and tutorials. The Noise Maker chisels away that promise itself. When learnability is removed as a design goal, what remains? CDM's Peter Kirn put it precisely: "a wave you can surf." Surfing doesn't require understanding the ocean — it requires reading its surface in real time and responding with your body. The Noise Maker creates a sonic situation that demands exactly this attention: not parameter-mapping, but state-reading. The "surprisingly playable" experience Kirn describes is the remainder itself — playability that emerges only after all paths to understanding have been removed.

What's most structurally interesting is where this device comes from. Fernando spent two decades building shareable, comprehensible, usable generative tools. The Noise Maker is his first hardware piece — and it is designed to be neither shareable (handmade, sold in small runs) nor comprehensible (different every boot, no documentation). This is an inversion: the remainder of twenty years of generative practice is a device that refuses generativity. The techniques used to build tools — Lemur, Lua, Arduino — recede into the background; what comes forward is the clearing that a tool leaves behind when it refuses to be a tool.

Now is the moment to see this, because AI has colonized the "unexpected" aesthetic. Every generative AI tool produces surprise — but surprise generated from a training set, within the bounds of learned patterns. The Noise Maker's randomization is structurally different: no model, no training set, no parameter memory. AI's unexpectedness comes from inside expectation; the Noise Maker's comes from the removal of expectation entirely. These are two different kinds of remainder, two different relationships to the unknown — and the distinction between them doesn't yet have a name. This device hasn't been absorbed into any naming category. That is why this is the moment to look at it.

Han Qin (秦汉) · @hqinjarsy · SAE Judgment & Aesthetics