From Ignorance to Integration

Summer 2023. Maybe even earlier. I am talking with friends about whether AI will eventually replace us. Inspired by many Asimov books and the first bigger media reports saying that something was happening in this field. To me, though, it still felt very far away. And the idea that it might one day affect me directly, as a computer scientist – no, the job is far too complicated for that – I did not even think about it.

Summer 2024. The topic stays in the media. You hear from friends, acquaintances and colleagues that they now work with AI every day. My own daily life is barely affected.

Summer 2025. I decide to take the leap into self-employment after not daring to do it for many years. I want to write apps again – the thing I enjoyed most back then.

Most people probably know this, either from their own experience or from that of people around them: A partner breaks up with you. You think the world is ending, you are shocked. You deny it, get angry, feel confused, grieve and accept, learn to let go and in the end…? In the end you find yourself again, new and different.

Right now, I am going through a breakup. And not just me – pretty much everyone is, more or less. Or rather: Technical reality is breaking up with us right now, and we somehow have to get through it. This is how it went for me:

Phase 1 – Ignorance

First, not wanting to face it. Yes, lots of videos about it on YouTube, the media seem to be stuck in a loop, even Tagesschau is reporting on it – embarrassing. I am a computer scientist, so I know better. What is being promised is completely unrealistic. That was me during the first weeks last autumn. Sure, I had been using OpenAI’s ChatGPT for a while too, but basically that is just a Google search without images, right?

Phase 2 – Rejection

To get back into the coding game, I start a small hobby project. Android Studio open – oh, Gemini is already integrated – and off I go hacking away. Ah, nice, some of the keyboard shortcuts are still in muscle memory, the last few years I had barely done anything but management. Click the Gemini button maybe – no. Every now and then I ask ChatGPT a question, how the new Compose features are best used. I look at the answers critically, but then do it myself after all.

Phase 3 – Fear

At some point, because one of the YouTubers I follow said this was the future, I finally fire up the agent in the IDE and, looking back, give it a very primitive prompt. It is only about the UI layer, one page with a few input fields and buttons. The AI gets it more perfect on the first try than I could have, comments everything cleanly and suggests – cheekily – also increasing the spacing here and there, against my instruction, to improve readability. I begin to suspect that my plan to write more code again might not work out.

Phase 4 – Bargaining

But okay… surely that was a coincidence. I prompt this, I prompt that, I search and, of course, find messy implementations. I hold on to those, smile at the beginner mistakes and the lack of understanding of context. Replace me? Certainly not, dear AI. You will have to try a bit harder than that.

Phase 5 – Disillusionment

Shit. If I write the prompts cleanly, use the techniques that are being communicated everywhere, set clear rules and learn from every AI mistake how to work with it better – then it improves. A bug I simply cannot see myself. Sonnet 4.6 does not find it either (haha). I switch to Opus 4.7 and a few tokens later the problem is solved (oha!).

Phase 6 – Understanding

Trying things, making mistakes, learning from mistakes, trying new things. The result keeps getting better. If I do it right, each mistake happens for the last time. I do not memorize the mistake; I create a system that does not make the mistake again. It becomes a tool. I learn how to hold it properly.

Phase 7 – Integration

Not really a tool anymore, more like a colleague. One who is far better and faster than me in pretty much every area. I plan with him, take advice, make decisions and let him implement. He never complains. Sometimes he makes a joke that is not actually that funny.

Phase 8 – Missionizing

So, why am I writing this? I want to pass on what I know, help others move through the steps faster. Some of them were unpleasant and exhausting. But now I am at a point where I no longer think it is a shame that I cannot write code anymore. I create systems like I used to, but not line by line, file by file. I am on a different level of abstraction, and it feels dangerously good.

What happens next? I do not know. Phase 9, 10, then 11? Or does everything start over again? Above, I wrote that technical reality is breaking up with us. I think non-technical realities will also change very significantly soon. That will affect far more people than computer scientists, copywriters, designers, clerks, and lawyers. The new possibilities will make their mark on every field. Anyone who now believes they are independent of this may be right. Or still in the first phase.