AI takes my code, what is left?

AI now writes code faster than I do. What a software developer is paid for is shifting from writing code to decisions, context and responsibility. According to Professor Pekka Abrahamsson, AI does in two minutes what used to take a developer two weeks.

I sit at my screen and watch the AI code: clean, functional, and considerably faster than anything I can manage. As late as February 2026, I was using it mainly as an autocomplete tool.

Still in the second half of 2025, I refused to believe that AI was going to upend very much at all. I had been refusing to believe it for a long time. I was also angry, particularly about how people seemed not to understand what large language models fundamentally are (more on that later too), yet imagined that AI “understands” what it is talking about. Then I told myself that sure, maybe it can write tests or handle trivial tasks.

Then I started using it seriously, in March 2026. Now I am simultaneously depressed and at peace with the situation. Depressed, because even though I had not built my (professional) identity around my coding skills, I am still proud of them. And now there is nothing particularly special about them anymore. At the same time, however, everything else I have learned over my career, the things programming has reinforced, is now more valuable than ever.

Why is the ability to write code now worthless?

Kent Beck, a pioneer of software development and a central figure in agile methodology, made a claim that stopped many in their tracks: “90% of my skills are now worth $0”. He too had concluded that AI has simply reached the point where writing code is no longer a domain in which humans hold an advantage. What once separated a senior developer from a junior, namely mastery of paradigms and syntax, fast code production and knowing the solutions, has gone. AI does it well enough. Sometimes it does it better, and it always does it in a fraction of the time a human takes.

Professor Pekka Abrahamsson of Tampere University, interviewed by Yle, sums up the change in concrete terms: AI performs in two minutes what used to take a developer two weeks. According to Metropolia’s Janne Salonen, a software project that used to cost 100,000 euros can in principle be done in 2026 for 25 euros. Etteplan and Vincit have already announced redundancies that the companies themselves linked to structural changes caused by AI. Nordea plans to cut an estimated 1,500 positions by the end of 2026, and AI and process optimisation are named as the explicit reasons.

This is the reality of software development in March 2026. I have since written in more detail about how AI changes what developers get paid for.

What can AI not take away?

Technological revolutions follow roughly the same pattern: first, something expensive becomes cheap; then people discover what the cheap thing now makes possible. According to Kent Beck, the number of things he can do expanded by a thousand. What he cannot delegate has proved more valuable than ever: knowing what questions are worth asking.

Andrew Murphy puts the same thought differently: “Software was never really about the typing.” Software development was never just about writing code. It was about understanding problems, navigating human dynamics, and recognising when a technically elegant solution is nevertheless the wrong one. Nolan Lawson describes the experience as grief, and that is the right word for it. The concrete satisfaction of cracking a problem in the code that has been nagging at you for days is gone now. And it is a real loss that should not be dismissed. But grieving does not mean you are required to stand still.

The skills that remain are not leftovers. They are the hard core of software development: the ability to make decisions, to read context, to understand responsibility, to make value judgements, and to know why the current direction was chosen in the first place. These are, in the end, the things AI cannot resolve on its own. I have written previously about AI’s limits in public services, and the same boundaries apply wherever responsibility remains human. My first newsletter reached the same conclusion: AI is not taking your job, but it is changing it.

How does the collapse in the cost of code affect society?

The biggest disruption will fall on mid-level developers, those whose skills consist mostly of producing code, implementing specified requirements, and hunting bugs. All things AI does faster, and (for now at least) more cheaply. Those at the beginning of their careers have probably been using AI from the start and use it more boldly and broadly than experienced developers. They now just have to learn what to supervise.

As the unit cost of code collapses, software production will industrialise. Suddenly even a small company can have applications built or build them in-house at reasonable cost, tailored precisely to what they need. The potential for new businesses and business models is substantial. Quality assurance, cybersecurity and related disciplines will of course become even more critical, particularly in sensitive systems. In the overall picture, however, those who do not adapt will fall behind. The same applies to the public sector: AI helps secure municipal services, but it is not free.

Just as AI accelerates processes within individual companies and projects, it will also accelerate a divide that was already beginning to emerge in society: the gap between high-productivity and low-productivity work. Put plainly: those who produce a great deal of value now will produce even more going forward. And those who produce less value will produce less still.

On a personal level, I am not afraid of the change, because I have for some time already been focusing on everything other than writing code. I believe there will be demand for my skills in the future.

The same cannot be said of society more broadly. I am not sure we are prepared for what happens when this same shift extends to knowledge work more widely. I have written about AI’s broader societal effects, particularly how the change is already visible in personal life and children’s daily experience. Software development is only the first sector to see these changes. It will certainly happen in other fields too. And there, too, people will have to work through the same thoughts.

The role of society is not, however, to prevent jobs from disappearing. Society’s role is to serve as a safety net for people when things do not go as planned. And that has never been more important than now.

If we assume things go reasonably well, that raises the next question: what would you do with the time AI frees up?