From Writing Code to Designing Systems: How the Developer Role Is Shifting
I made a living writing good code for over fifteen years. Today, a large share of that code is written by an agent — faster than I ever could. I asked myself the obvious question, and it was uncomfortable: what is left of my profession?
The short answer: more than the headlines claim. But in a different place than where many developers stand today.
The shift, concretely
Vibe-coding is real. You describe what you want and get a working app in minutes. For prototypes and internal tools, that is a breakthrough. But look closer and you also see what gets dropped: architecture decisions, test coverage, review, security, maintainability — the last twenty percent that turn a demo into a product you can sell and operate.
Those twenty percent don't disappear. They become the core of the job.
Yesterday's developer was paid to translate requirements into code. Today's AI systems expert is paid to decide which code should exist at all — and to make sure what gets built holds. That sounds like a small shift. It is a fundamental one.
What loses value — and what gains it
Losing value: syntax knowledge, framework trivia, the ability to type a lot of code quickly. Anything an agent can recall from its training is no longer a differentiator.
Gaining value: architectural judgment. The ability to break a system into verifiable units. Recognizing when a plausible solution is wrong. Specifying requirements so precisely that an agent — or a human — cannot misread them. And the hard-won experience of knowing where software breaks in practice.
The irony: exactly the senior skills that the vibe-coding hype treats as dispensable are becoming the bottleneck. Agents produce code in abundance. What is scarce is the judgment that decides whether that code fits into a system that will still be maintainable in three years.
What that means in practice
If you are a developer wondering where this is heading, I would offer three things:
- Learn to work with agents, not against them. Integrate the tools seriously into your daily work — not as a toy, but as part of the workflow. The learning curve is real, and right now it is the best career investment available.
- Deliberately move toward the last twenty percent. Review, architecture, testing, operations. These are the skills getting more expensive, not cheaper.
- Stay close to the business. The most valuable question is no longer "how do I build this?" but "which of this is worth building?". If you understand both sides, you are irreplaceable for the foreseeable future.
The profession is not disappearing. It is becoming more demanding — and honestly, more interesting. The typing was never the most valuable part.
The long version of this assessment — including my own story from platform architect to AI systems architect — is available as an audiobook: Vom Entwickler zum KI-System-Experten (published in German).