in progress!
I’m not predicting anything.
No collapse, no hard times, no apocalypse. There are plenty of others for that. What I see is more banal and more urgent at the same time:
We have to rethink faster now than we actually can.
Technology changes faster than people can make meaning out of it. That’s not a threat, it’s an observation. And it’s the only reason I still write, build, disrupt.
Three sentences that carry everything
Technology serves people, not the other way around. If a tool demands that people bend their routines, their relationships, their identity to fit it, it’s badly built. The human isn’t deficient. The tool is.
AI needs its own operating system. Today’s models run on infrastructure built for other purposes. They’re docked, not at home. Sovereign AI means: local inference, your own data, your own key, no dependence on corporations that could be different companies in five years.
Build for 2030, not for 2026. Whoever works today with technology that’s obsolete in four years is building debt. Whoever works with technology that will be standard in four years is building capital. Big companies can no longer make that jump — too much inertia, too much legacy, too many committees. Solo builders and small structures can. That’s the only advantage they have — and at the same time their entire leverage.
How I work
Local where possible. Inference on my desk, not on someone else’s servers. Data under my control. If cloud, then sovereign infrastructure, German metal, my own racks.
With several models, not just one. One model writes, another checks, a third attacks. Whoever works with a single model thinks from a single standpoint — that’s not thinking, that’s repeating.
Transparent about co-authorship. If an AI helped write something, it says so below. Which model, in which role. No hidden synthesis. I don’t claim what I didn’t do alone.
Function before brand. First it has to do what it’s supposed to do. Then it gets to have a name. Not the other way around.
From maker to observer. Sixteen years running my own company, eleven years at TIMOCOM before that, IT since 1998 before that. In 2025 I shut down Webbinder, because I realized: now what counts isn’t what I produce, but what I understand. Building without a brief, observing without comment, writing only when there’s no other way.
Why extremes sometimes
I reach for means that touch people. Not to shock, but because reasoned arguments sink into a real-time feed before they can take effect.
A pastiche of H.G. Wells, a digital Frankenstein experiment on LinkedIn, a blunt vulnerability disclosure to the BSI — these aren’t provocations for their own sake. They’re attempts to reach a level of discourse where a person still asks instead of scrolling on.
If you get the impression I’m overdoing it: that’s the intent. A person who’s been shaken awake can ask. A tired person clicks on. Between those two states is where it gets decided how the next few years go.
What I don’t do
I don’t sell AI. I don’t sell consulting. I don’t take briefs in the classic sense.
I don’t preach collapse. I don’t claim we’re lost. That would be just as cheap as claiming everything will be fine.
I don’t optimize for reach. I write little, I post little. What stands here is meant to last, not to trend.
I don’t go to people. I let people come.
I don’t run along with the hype. When everyone uses a tool, I ask about the why, not the whether.
What I do
I build systems that can hold up in four years — local, decentralized, post-quantum-ready. I write texts that three people will say helped them. I try not to fall off the bridge between code and art, because both sides are needed.
At the center stands logpy — the backbone everything else docks onto: MARTIN, LogpyClaw, ALICE, OpenNet. A different answer to the AI question than “faster, bigger, more corporation”. The names say little, the function counts.
As always: three to five years too early. I’ve gotten used to that.
I don’t sell AI. I live with it.
If you have an idea and it holds up against the compass of values, I’ll build with you. If not, I’ll tell you so.
— written down with Claude Opus 4.7 & H.G.O.D. Memory, June 2026
Leave a Reply