The ideas were
always there.

01 / THE IDEAS WERE THERE

I could picture the product.
I just couldn’t build it myself.

Around 2010, the internet still felt like an open invitation. E-commerce. New business models. Entire industries waiting for someone to ask a better question. My head was full of products I wanted to make.

Could machine learning help us make better credit decisions? Could I build a prediction market — the kind of idea people now associate with Polymarket — where expectations about the future become something you can actually see?

The ideas weren’t the problem. The distance between an idea and a working product was. I needed people who could build what I imagined. That meant translating, coordinating, waiting, compromising. Each attempt came with an organisational bill.

Building with people can be brilliant. But depending on a team for every experiment put a ceiling on my own creative freedom. I couldn’t simply follow a thought, try it, dislike it, and try again until it felt right.

I had the picture in my head.
I didn’t have a direct route
to making it real.

02 / MY AHA MOMENT

Now I can just
start building.

That is why this moment hits me so hard. Working intensively with Astra and Codex has changed the distance between thinking and making. I can describe what I want in a voice message, look at the first version, explain what feels wrong, and keep going.

I don’t need to assemble a team before I can attempt something. I can sit comfortably at my computer and work on a product that, a few minutes earlier, existed only in my head. That immediacy is the breakthrough for me.

I’m talking about more than a chat box that produces a clever answer. In my workflow, agents work through consoles, write code, handle files, and use tools. Several can work on different pieces at once. I set the direction, inspect what comes back, and steer the next pass.

It doesn’t make every idea a good idea. It makes ideas testable. I get to discover what works by actually making it, in my own rhythm, with my own taste. For someone who spent years carrying products around in his head, that is a pretty extraordinary upgrade.

03 / YOU ARE STILL THE PILOT

An incredible spaceship.
Still needs someone at the controls.

Extraordinary range.
Your choice of direction.

This distinction matters to me: today’s AI feels like an extremely capable, manually piloted aircraft. Or a spaceship, if Rosi gets to pick. It has astonishing power. How well it flies still depends heavily on the person at the controls.

What are we actually trying to build? Who is it for? Which information does the model need? How do I phrase the prompt, organise the console work, and split the tasks? Most importantly: can I tell when an impressive result misses the point?

Speed doesn’t fix a vague brief. It can just turn a vague brief into the wrong thing very quickly. Clear prompts, useful context, and honest feedback are not finishing touches. They are the steering.

AI brings the thrust.
The pilot holds the course.

This website is part of that learning process. It shows what happens when I get properly stuck into the tools, follow my curiosity, and keep caring about whether the result is any good for the person using it.

04 / THE DOOR OPENS WIDER

The next good idea
doesn’t need a whole department.

A small tool. A peculiar website. A solution to a problem nobody else considered worth solving. More people can now turn those thoughts into something real without first raising money, starting a company, or hiring a development team.

A computer, an internet connection, access to capable AI tools, and a willingness to learn can be enough to begin. That feels like an enormous opportunity. A useful idea should not die because its owner cannot afford a software department.

It doesn’t magically give everyone equal starting conditions. Time, language, knowledge, access, and ongoing costs still matter. An AI subscription is not a finished product. But the door is open to many more people — and I want to see what their odd, specific, ambitious ideas look like on the other side.

05 / LOVE, PASSION & AI

When building gets cheaper,
care becomes more valuable.

My bet: the effort required to implement software will keep falling. As that happens, simply being able to build something becomes less of a distinction.

What are you building? For whom? Why this way? Which detail refuses to leave you alone? When do you say, “Yes, it works. But it doesn’t feel right yet”?

That is where the love lives. In a sentence that sounds like a person. A button you understand without thinking. A character like Rosi that turns a utility into something with a personality. In looking one more time after the first version was already technically good enough.

AI helps me build more.
I still have to bring
something worth caring about.

That is the combination I’m interested in. Useful things with a point of view. Products that feel considered, because somebody was genuinely interested in how it would feel to use them.

06 / AND TOMORROW?

One day, the operator
may become optional too.

Right now, being good at steering AI is an advantage. Explaining a task, seeing the connections, assessing results, knowing when to intervene: that is still work. It is also something I’m enjoying learning enormously.

But how long does that advantage last? As systems become better at planning, using feedback, and checking their work, they may take over more of the steering too.

The date interests me less than the question behind it: if we are needed less and less to execute things, what do we want to spend our time on? What makes us curious? Who do we want around us? What would we build even if nobody paid us?

I don’t have a final answer. For now, I want to use this strange, exciting window to make things that fascinate me. Build good projects. Work with people who care about their idea. And remember that life also happens outside a browser tab.