How I think

Plot twist. This website was built with AI. Yes. I see the irony.

I'm not a software engineer. AI helps me build things I couldn't build on my own. But someone still has to understand the problem, question the answers, make decisions, test what works and figure out what to do when it doesn't.

That's the part I enjoy most.

Three true stories. Not to show you what I built, but how I got there. Mistakes included.

CASE 01

How to save €1,500. The expensive way.

€20

A salesman wanted €1,500 for a new phone system, plus a yearly contract for maintenance and support. I built my own instead. Four months later it worked, and it stayed on for thirteen years. I paid the real price later.

  1. The problem

    The old analog phone system still worked, but it wasn't enough anymore. A salesman kept calling: €1,500, new phones, and then a contract, year after year.

  2. The choice

    I didn't want lock-in. I wanted to understand. So I started: evening after evening, night after night. And, I'll admit it, a few working days too.

  3. What went wrong

    Four months. The system kept freezing, and customers called to say the phone wasn't working. I was even scared to buy a single VoIP phone, just to save money.

  4. The result

    A recycled motherboard, a €20 no-name power supply, two ISDN cards and Linux. It was meant to be a test. It ran in production for thirteen years.

  5. The real price

    If the phone system broke, they needed me. If a server went down, they needed me. I'd saved €1,500, but I'd made myself indispensable. And I couldn't hand my precious over to anyone.

I didn't want lock-in. I became the lock-in.

CASE 02

An app to find out if I could stop working. It had me working twelve hours a day.

I wanted a system to tell me whether my investments could cover my expenses, so I wouldn't need to work. I kept adding pieces, with AI helping me build them. Six months later it was so complex that a small change took hours. I decided to rebuild it from scratch.

  1. The problem

    Investments spread across several accounts, everyday expenses, and one question: could I live without working? I'm not part of the FIRE movement, but the question intrigued me.

  2. The choice

    Build it myself, one block at a time: expenses, then investments, then everything together. Then AI to read receipts and bank statements.

  3. What went wrong

    Every piece worked. The whole thing didn't. Six months, twelve-hour days, paid subscriptions, and in the end a simple change took hours.

  4. The decision

    Stop, and rebuild the back end from scratch with one clear goal: each user's data encrypted with their own password, so not even the admin can read it.

  5. The lesson

    With AI, adding things is easy. Deciding what not to add is still your job.

My wife called it in month three. She was right. She usually is.

CASE 03

The graphics card I sent back.

The new GPU, just arrived, still in its plastic wrap, in its box.
GPU unboxing

I wanted AI at home, without sending my data to anyone. I bought a graphics card, got cold feet and sent it back. A week later, prices were up 50%. I bought one again. A bigger one.

  1. The problem

    I was building a system to use several AI models together. It worked. But all my information was leaving my house.

  2. The choice

    Weeks spent studying graphics cards and running tests. Then the first card, with 16 GB.

  3. What went wrong

    I panicked: I can't pull this off. I sent it back. A week later, it cost 50% more.

  4. The result

    I worked up the nerve and bought the 32 GB one, for €2,000. I'm still not sure it was the right call.

  5. The reasoning

    A subscription costs €20, €30, €50 a month, and in two years that card will be outdated. But keeping my data in my own pocket, instead of handing it to people I don't know, is worth more. That data is our life.

In two years it'll be outdated. My data will still be mine.

KNOWN BUGS — NOT FIXED YET

  • I get excited. I move fast. Sometimes too fast.
  • Then I overthink, get anxious, or simply run out of energy.
  • Still working on my battery management system.
  • That's why I bought an electric car.

I don't solve unsolvable problems. But for the solvable ones, I'll do everything I can.

Let's talk