Background

The job was never the language.

In becoming a programmer I learned programming languages in order for systems to function and communicate with their users. In some ways it seemed like the languages were what it was all about. Removing the languages made it quite clear what the job was really about.

I don't write JavaScript any more, or C#, or SQL. The models write all of it now, and that is fine, because it allows me to focus on the important parts: The functionality and how users interact with it. The language played its part but the more important skill was working out what a business actually needs, building the thing it then runs on, and knowing when software is not finished regardless of what the calendar says. That part has not been automated. It is what the years since 2000 were spent learning.

What I have built

I started in the summer of 2000, at a temp agency in Copenhagen that was building a platform. I built it: the website and the back end, a small system for managing clients and matching them with applicants, and hour registration for everyone placed on temporary contracts. It covered nearly everything the agency needed, short of the accounts. I was the only technical hire in the early days, and I learned most of what I needed to do on the job.

I left what felt like a safe and caring place in favor of a company that sold digital signage. Here I built another platform and customized solutions. This included systems for displaying advertisement, film and interactive content on large screens and information kiosks, for shops and businesses. Its customers included a Danish cinema chain, a major phone company, a real estate chain, a food chain and Copenhagen airport. The company sold the hardware along with the software, which doubles the ways a thing can go wrong. Among a host of new emerging technologies I remember in particular how I learned to intercept keyboard input in C++ so that nobody could close the program on a kiosk. There was far more software to write than time to write it, and some of it was sold before it was built. That is the part I would not repeat, and the reason I can tell when it is happening again.

After that I moved to Lithuania and started an IT company there, building bespoke software for Danish clients. The proposition was a gap in the market rather than a sales pitch: cheaper than a Danish agency, and near enough to be reachable, unlike outsourcing on the other side of the world. I went from writing the software to running the company, which is a different job and, as it turned out, not the one I thrived in.

Since then I have worked alone, or leading small teams: fewer clients, done properly. For several years I built the first systems for a blockchain company, then hired and led its development team, mostly in the Philippines, and did the testing myself. Along the way I also built a marketplace platform that never found its footing, in part because Facebook Marketplace launched at the same time; I have done experiments on the side throughout the years. Two of the promising ones can be found on these pages.

One gig has lasted fourteen years

For about fourteen years, a lot of my time has gone into one system, for a Danish company in the construction business.

Their sales and technical staff use it to put together the solutions they sell: what each one will cost to produce, what has to be ordered, and how it is built. It keeps growing because their product range keeps growing, and there is always something new to add. I maintain their main website and several smaller ones as well.

Fourteen years on one system teaches you something a shorter run cannot: you are still there when the decisions you made at the start come due. You learn which corners can be cut and which cannot, because you are the one who comes back and mends them. That is where a lot of what is on the Kernel page comes from.

Many JavaScript frameworks have come and gone in that time. It was never built on the one everybody was sure about, and it still stands.

What I do now

The same judgement, with new instruments. I direct AI agents instead of writing every line, which is closer to leading a development team than to typing: deciding what is worth building, seeing which of several approaches is the right one, reviewing what comes back, and testing it properly before anyone else sees it. That is not a new job for me. I have built the first version of a system, then hired a team spread across the world to carry it on, reviewed what they sent back and tested it myself. The instruments are new; the work is the one I have done for years.

I am older than when I started. There are reading glasses now, and less hair. But I am sharper about what matters, and with these tools I build working software faster than at any point in my career.

That is also what makes me useful to a business built before any of this existed. I am not arriving to tell you that everything you have is obsolete and should be replaced. Much of what companies run on was built carefully, by people who knew what they were doing, and it is still holding. The interesting work is in what these tools can add to it.

Where to go from here