So, I would say: for a long time I only built. Writing felt like time I did not have, and like an opinion nobody had ordered. That has changed. Not because I suddenly feel like posting. But because I noticed that for years I have been writing down what I learn, and then it sits in a drawer. It is no use to anyone there. And right now quite a few people are sitting in front of the same question I asked myself.
For a long time I only built
Since childhood I have been fascinated by electronics and software. Back then I was allowed to learn both, which was not a given, and I stuck with it. If you read this blog from the back, you see what came of it. A fridge that knew what was inside. A model that recognised plants in the garden. An app that scans rooms. A museum in the browser. A catapult game where we calculated angles for a year. Plus order processes, test data, small models, agents that are not allowed to sign off on their own work. None of it was a big throw. It was workshop: make, fail, make again, ship. And then the next thing.
I never wrote about it. I thought whoever builds does not need to talk. And I thought that those who talk a lot have mostly built little.
The second I still think. On the first I was wrong.
The notebooks
Because I did write, just for myself. For many years, with pen and paper, in small squared notebooks, with sketches in between. No tool, no app. Not what you imagine under a diary, no feelings, no weather. But the sentence that fell in a call. What was on the whiteboard before it was wiped. Why a model did not work although the numbers looked good. What a customer said shortly before a project tipped. How a team looked when the first agent wrote code.
Last year I started leafing through the pile. There are more notebooks than I thought. And while sorting I noticed something: most sentences are not about technology. They are about decisions. When to start. What to rebuild first. Who to bring on board. What to tell a team before it gets restless. And almost every one of those sentences I would have liked to hear ten years earlier from someone who already had it behind them.
It is coming anyway
And here is the real reason I am writing now.
I am sure: AI is coming to almost every company. Not as a chatbot on the website, but where it hurts and where it pays: in the processes and in development. Orders an agent checks before a human sees them. Code that to a large part is no longer written by hand. Documents that are read and sorted without anyone opening them. I have seen it for two years in every project, at big houses and at companies with twenty people.
And I see something else: most are not prepared. Not because they could not be. But because nobody told them what it means concretely. That the data has to be clean before a model can do anything with it. That a team with an agent needs different rules than one without. That you start with a process, not with a strategy. The difference between the companies that get through this well and the others is not the technology. It is the preparation. And it is not complicated, you just have to do it.
Both sides of the table
What I can contribute is not a recipe. It is a point of view, and it comes from having sat on both sides of the table.
Most people who help companies with AI come from one side. From consulting: then there are slides, a target picture, and nobody in the room who knows whether the thing can be built that way at all. Or from technology: then there is a model that works, and nobody who knows how it gets into a process that twenty people have been doing the same way for ten years. Each side lacks the other.
I have done both, not especially elegantly, but long enough. I have built models and written tracking and set up agents. And I run my own companies, in Switzerland and in Dubai, with everything that comes with it: teams that did not want a change, budgets nobody understood, projects that went well and ones that did not. I know how a model is converted, and I know what it feels like when a team is two months slower after a change before it gets faster.
Who I write for
For people who are standing in front of the same thing right now. Three groups, concretely.
Companies that know it is coming and do not know where to start. Twenty to two hundred people, a core business that runs, and the question: where do we start without breaking everything? That is exactly the question I had to ask myself with my own companies.
Teams that got an agent and do not know what now. Developers who suddenly review instead of write. A lead who does not know how to lead when half the code no longer comes by hand. I went through that with my own teams, and it works if you clear up a few things beforehand.
People with an idea, before building. Founders, developers with a free weekend, someone from a completely different industry who has seen something. Before building, when you can still decide what comes first. That is where one sentence from someone who has already tried is worth the most.
What all three lack is mostly not motivation and not money either. It is someone who says: I did that, this is how it went, and this is what I would do differently today. Not as a lecture. As a conversation between colleagues, with your own thing on the table.
What you find here, and what not
What I built, and what went wrong. The projects from back then, told honestly, with what I would do differently today. And the things from now: small models, agents, test data, teams, as they happen.
My opinion, marked as opinion. Where I can back something up, the source is there. Where it is experience, it says experience. Where I am only sure, it says that too. I try not to mix them.
No tool lists, no hype. No episode about AGI in which I pretend to know when it comes. No ten tools you have to know this week.
No selling. This is my private site. What my company does is on its own. If I mention something of ours here, it is because the article is about exactly that, and then I say so.
How this blog comes about
This is not a content plan. It is the notebooks, only that others may read along now. Unfinished, concrete, no gloss. Irregular, because I build during the day.
And because the question comes anyway: yes, a model helps here. Our own, a twin model trained on my texts. It corrects grammar and spelling, nothing more. The thoughts, the values, the learnings, the mistakes, the projects that went wrong: all of that is mine. No model built SKIMO, calculated catapults for a year or sat in Dubai in February. Sometimes it turns into a mix. Of course I read everything again, but nobody is infallible. Whatever stays in is mine too. That is written under every article, so it is clear before anyone asks.
Whoever only posts without having built has nothing to say. Whoever only builds without telling helps nobody who comes after. I am now trying both. Not because I know better. But because I have already tried a few things, and that can save you a week.
For a long time I only built and wrote for myself. Now I write for those who want to be prepared when AI comes into their processes and their development. From the workshop, without slides, from someone who does the same thing during the day as you. If you are standing in front of your own thing and want to talk: the "Meet" page says how. That is the reason for this blog.
Written by me. The thoughts, the values, the learnings, the mistakes: all mine. Grammar and spelling are corrected by a twin model of my own, trained on my texts. Sometimes that turns into a mix. Of course I read everything again, but nobody is infallible. Whatever stays in is mine too.