Who I am,
in interests.

Orchestrating rules and laws, without obligation.

Composing and producing is systems thinking: harmony, rhythm and arrangement are rule sets you know in order to break them deliberately, and the signal chain underneath (gain staging) is pure engineering.

There's a second layer that weighs just as heavily — an empathic mental model of the intended listener. That coincides directly with preparing a DJ set: I weigh every track choice up front against that abstract mental image — where do I want to move the listener, and what gets them there.

I'm the intuitive operator of a fundamental algorithm.

I make algorithmic art; generative art is one branch of it. My primary work: using a cash-register keypad I apply convolution by feel — sharpening, blurring — until arresting abstract images emerge.

These are the conventional algorithms — everywhere and running constantly as of 2026; I play them by hand.

Capturing a state — artificial or natural.

For me photography is capturing a state, with as few means as possible.

With a Canon EOS 700D and a 10-18 I shoot better event images than the photographer next to me with a Sony A7S III; studio product shots I take with a Sony MVC-FD91, whose 640×480 output ironically needs no web optimization at all in a Next.js app and is visually indistinguishable from a modern photo.

And after the shoot: paring a heap of photos down to a tight series.

Bringing the digital into the real world, pixel-perfect.

With modified inkjets (Epson R2100/R2400) I bring my work physically into the world. The technique grew out of a requirement: my art (64×64 to 512×512px) has to come out pixel-perfect and crisp at scale.

Where a conventional print process picks bilinear, my work actually needs nearest-neighbor for sharp pixel edges. On top of that the whole color-profile domain — for me a requirement rather than a passion — but it is fun to push cheap ink and media to their theoretical limit.

My measuring instrument is a GretagMacbeth i1Pro that I bought as “no longer usable” and, with ArgyllCMS and Gutenprint, got to deliver Swiss top quality again.

A system you can fully see and measure.

Analog electronics (and the audio that comes out of it) is fun because the low-level system is almost fully visible and measurable: no spec sheets needed to understand or troubleshoot it.

It sits closer to mechanics than to digital, and often goes hand in hand with ingenious physical mechanisms that solve in hardware what a digital system would do with code — the insides of a cassette deck that, before the bi-head era, auto-reverses the tape via a giant roundabout mechanism, or playing with the constant 50Hz.

The machine that produces on its own for days on end.

As an apprentice machinist I programmed, set up and operated CNC machines and automation robots. The real work is the program that lets the machine produce autonomously and with as little downtime as possible.

That means designing for what goes wrong: tool wear you predict and compensate for, fallback patterns when something drifts, the right program and strategy choice for the run, and measuring in the loop. And at scale every decision adds up — one second per cycle is hours over a run of 10,000 pieces.

A tool within reach, not a love in itself.

CAD and 3D printing are more tools to me than an intrinsic interest.

It's fun to get the maximum out of the printer — in line with the inkjet calibration, a costly and linear process — but in practice it's simply the fastest route to a prototype or a replacement part, with minimal turnaround.

Respect the trajectory.

In acro mode the drone doesn't stop when the throttle closes — the trajectory keeps going, and for even a 'simple' turn I manipulate all axes at once.

It's full motor immersion: thinking ahead along the path, constantly correcting, and in that flow state a sense of freedom that few other things give me.

First understand why, only then the screwdriver.

My default is intuitive: grab a screwdriver and fix things by feel, which almost always works. I only follow a procedure when the stakes are high (expensive, not my own product — like at Coolblue) or when it isn't solvable straight away.

It's more about the cause-and-effect thinking and understanding the system I'm repairing than about completing a routine — which, by the way, is often just a replacement (screen, battery), which strictly speaking isn't a repair.

The brain is just an unsteered stream of thought too.

What really fascinates me is the overlap between how a brain works and how an LLM or agent works. The stream of thought you don't fully control resembles a model's generation; and the brain few-shot-prompts itself from a feeling — a depression loop is in effect a self-reinforcing prompt that colors everything that follows.

That I see that overlap is exactly what I find interesting about it.

The most compact unit of communication there is.

I'm crazy about metaphors. With no particular prompting I find myself building a strong analogy, with the bar being: my mother has to get it. It's about compressing something complicated into the smallest possible unit that conveys it.

I discovered it in tech support (MediaMarkt, Coolblue/Apple), where I had to explain to laypeople why their problem isn't covered under warranty — high stakes, and almost always explainable provided you truly understand it yourself.

Putting reality in perspective and in place.

For me astronomy is mostly about perspective: how the (im)balance across the whole horizon is what put me here, and — with a generous over-generalization — that I'm describing that very thing by way of an entity that itself arose entirely from natural raw materials. Systems thinking at the largest level.

Spaceflight follows from that: discovering that system, plus the technical marvel that a craft stays controllable at barely comprehensible distances — especially in the early years, when it was pulled off with minimal means and pioneering knowledge, on computing power my DMG-01 Game Boy would beat.

Where all the threads come together.

Since mid-2025 I've been building full-time: full-stack in TypeScript and Python, terminal-first with Claude Code and multiple models. It feels like music, metal, image and language compressed into one medium — the same instincts, in a new material.

Finding the rules, designing the system, building it, reviewing critically, repairing until it's right, and explaining what's happening.