A place for poietic work

What I want to build around Poieticism, and who I hope builds it with me


poieticism community making

When I was a kid, nobody had to explain Lego Technic to me. The box had parts, the parts fit together, and when my tractor fell apart it was clear whose fault it was. Kids today also get a different box: a text field that returns a finished tractor in a few seconds.

I keep thinking about what it would take for them to treat AI the way I treated that box, as a set of parts that fit together and sometimes fall apart in instructive ways. That question is where the posts about the word and the first principles were heading, and in this one I want to write down where I’d like Poieticism to go, even if it sounds too big.

The big version

In ten years, I’d like poietic work to be something people recognize and practise on purpose, the way people practise open source today. Somebody makes a thing and puts it in front of the world. They write down what pushed back and what they changed, they keep the thing running or retire it properly, and the next person ends up more able to make the next thing.

I’d like there to be places for this in many cities. Hackerspaces, university labs and school workshops already exist, and next to them there could be virtual rooms where people from different countries build together. In all of those places, AI would be the most useful tool on the table, used in a way that makes the people holding it better at making.

I know how this sounds. I’m one person with a blog, a simulation company and a WebXR app.

Starting with my own tools

The part I fully control is what I build, so that’s where it starts.

  • In Graspable, the agent edits a real project folder on your computer, which you own and can open in any editor. A run only counts as done when the project builds and loads in a browser, and anything risky waits for your approval. Looking at it now, that’s the second and sixth step from my list turned into software: the browser gets to push back before the agent says “done”, and the result stays yours.
  • At DimensionLab we started with Siml.ai, and today we build Frontier. It takes the simulations an engineering team has already run, trains physics models on them, and shows on the geometry where each model is wrong. Simulation lets physics push back on an idea before anybody spends material on it, and a model that marks its own errors keeps that push-back in view.
  • On this blog, I’ll keep writing and tag every post about this with poieticism, so it’s easy to find in one place.

The experiment I’m running

The boldest part is something I’ve already started. At DimensionLab I’m working on a project called Supersolve (not public yet), where AI agents work on physics problems on their own and keep improving how they work. The longer plan is that the agents gradually take over DimensionLab itself.

I’d stay in the company as a meat proxy: the human the agents can send out whenever something needs a person in the physical world. At first I’ll be the one deciding what that work is. Later, I want the agents to hand me tasks for whatever they need. Yes, I’m volunteering to become the hands of my own software, and my job description will be the strangest one in the company.

From the poietic side, this is the real test of the first principles. I wrote that making has two results, the thing and a changed maker, and that the system doing the searching is the one that learns. Here I’m leaning into it on purpose, because the agents are the makers who get changed. The second step still applies to them: nothing is made until the world pushes back, and for agents working on physics, a simulation checked against another simulation isn’t enough. So part of my human work will be building physical models, real objects I can measure, which the agents will use as validation datasets. The physical world gets to push back on them, through me.

Agents running a company need clear limits on what they can decide and spend. I wrote about that in An agent needs a boundary and When software can pay, and now I’ll find out whether I meant it.

People

From running QuestSpace and being part of Paralelná Polis Košice, I know that ideas get real when there’s a room, some tools and people who keep showing up. A blog post can’t do that part.

So consider this an invitation. If you make things, with AI or without it, and you care about the changed maker as much as about the thing, I’d love to hear from you (there’s an email link on the front page). Send me what you made, and above all tell me what pushed back. I’d like to collect these, and once there are enough of them, meet in person and see whether a community grows out of it.

There won’t be a manifesto with a logo. If Poieticism ever gets a written set of principles, they will be short, because I already argued that whoever makes a thing has to maintain it, and I’d like to keep my weekends. 🙂