Different Materials. Same Instinct: Build.

For a long time, I treated the different parts of my life as separate things.

There was work: software, cloud, AI, products, architecture, teams, companies.

Then there was everything else: HYROX, smart-home systems, tools, LEGO, fixing things around the house, taking something apart because I needed to understand why it worked that way.

They looked unrelated on paper.

They are not.

Different materials. Same instinct.

The common thread is building.

Not just making things for the sake of making them. I mean the full loop: understand the system, form a theory, build something real, collide with reality, measure what happened, then make it better.

That is the same pattern whether the material is code, a product, a training plan, a KNX panel, a broken shutter, a piece of furniture, or a business.

I am usually happiest somewhere in the middle of that loop.

I never really stopped being an engineer

My education is in software engineering, but the useful part of engineering was never the programming language.

It was learning to decompose a problem.

What are the parts? How do they interact? Where is the bottleneck? What assumptions am I making? What can I test quickly? What breaks first?

Those questions follow me everywhere.

In AI, that might mean moving past an impressive demo and asking whether an agentic system can survive a real enterprise workflow.

In a house, it might mean understanding the relationship between KNX, Home Assistant, sensors, shutters, HVAC and networking instead of treating each device as an isolated gadget.

In training, it might mean looking at pacing, lactate, running economy, station efficiency and recovery as one system rather than a pile of workouts.

Even assembling IKEA scratches the same itch. There is a system, there are constraints, and there is something satisfying about making the pieces become the thing.

Build first. Opinion second.

I have become increasingly suspicious of opinions formed too far away from the thing itself.

You can talk about AI architecture for weeks. Then you build the workflow and discover the hard part was state, permissions, latency, tool reliability or the human checkpoint everyone forgot to model.

You can design the perfect training plan. Then you race and learn that the transition you ignored costs more than the fitness gain you obsessed over.

You can specify a smart home beautifully on paper. Then a sensor behaves differently in the actual room and the whole automation feels wrong.

Reality is a very efficient reviewer.

That is why I prefer prototypes to certainty.

Measure what changes a decision

I like data, but not as decoration.

A dashboard that never changes a decision is just wallpaper.

The useful metric is the one that makes you do something differently.

That might be a product usage signal that kills a feature. A race split that exposes bad pacing. A recovery trend that changes training. A system log that points to the real bottleneck. A sensor reading that tells you the automation is wrong.

Measure. Decide. Adjust.

The loop matters more than the metric.

The subjects will keep changing

I do not expect this site to stay in one lane.

Some weeks it will be about AI agents or product architecture. Other weeks it might be HYROX, a mobile app, a smart-home experiment, a tool, a repair, or something I built because I could not leave the problem alone.

I am okay with that now.

The subject is not the identity.

The behavior is.

Understand the system. Build the thing. Test it. Measure it. Improve it.

Different materials. Same instinct.