The developer's craft has been dying for 40 years. And it has never been healthier.

I’ve been hesitating to write this post for a good year. Not because the subject is hard, but because it departs from my habits here. Usually, I share things you can verify: measurements, protocols, lessons from experience. Here, I’m simply going to give an opinion. Owned as such.

A colleague shared an article this week about the identity crisis of software engineers in the face of AI. My first reaction was: “I know exactly who feels that way around me.” My second, more honest one: “for me, it’s the exact opposite.”

For two years now, I’ve known that the craft, as we used to practice it, is dying. Two years is more than enough time to accept and rebuild. So the wave of anxiety arriving today, I’m not living it the way those in the middle of it are. I’ve crossed over to the other side.

But careful. It’s easy, once you’ve digested something, to look at those still digesting it with a touch of condescension. That’s not my intention. What’s happening is real, and the suffering is legitimate.

The craft has been constantly dying for 40 years

A bit of perspective.

I learned Merise around 1996. I was about ten. For my mother back then, coding without doing Merise was unthinkable, so she handed me all her books on the subject. I still keep, in my desk drawer, her template ruler: a little stencil full of cut-out shapes for drawing diagrams by hand. I haven’t used it in twenty years. But it’s there. Then Merise vanished, replaced by UML. I coded in Basic. Then UML vanished. Then Basic vanished. The open web arrived and changed everything, then started dying too, eaten away by the platforms.

I could go on for a long time. Hand-administered servers, swept away by the cloud. The monolith, buried by microservices, then half-exhumed once we saw their price tag.

For 40 years, our craft has been dying and being reborn on a loop. With every cycle, a generation watches what it had mastered lose its value. And every time, it’s right about one thing: that particular skill, yes, does lose value. AI is the most violent cycle I’ve ever seen. But it’s a cycle.

The loop doesn’t move forward. It moves back.

There’s a detail in this story that I find important.

In the early days of enterprise computing, the work was split into two jobs. On one side the analyst, who understood the need and designed the solution. On the other the programmer, who translated that solution into code. Two people, two functions: think, then execute. A very Taylorist organization.

Then, over the 1960s-70s, the two merged into a single role: the “analyst-programmer,” the direct ancestor of the developer we know. The person who designs is the one who codes. It’s into this already-merged craft that my mother entered, and it’s the one I inherited.

Now look at what’s happening today. AI re-splits exactly those two jobs. On one side, someone who understands the problem, decides, designs, validates: the analyst. On the other, an entity that produces code from intent: the programmer. Except this time, the programmer is no longer human.

The loop doesn’t just turn. Sometimes, it returns to a state we thought was behind us. We had merged design and execution; we’re now re-splitting them. And that’s precisely what’s disorienting: what we’re living isn’t an absolute novelty, it’s an old boundary reappearing.

In the very beginning, programming was largely women’s work, before becoming heavily masculinized between the 1960s and 1980s. The boundaries of the craft don’t only shift on the technical plane. They also shift around who is allowed to practice it. That would deserve a post of its own.

What’s dying isn’t what you think

Saying “the developer’s craft is dying” is false. It’s not software engineering that’s dying, it’s one way of practicing it: writing lots of code by hand, mastering a framework as a differentiator, being valued for raw production velocity.

That, yes, loses enormous value. And that’s the real subject of the anxiety. Not “I’m going to lose my job tomorrow,” but: “the thing I was proud of, the thing that set me apart, a machine now does it better than me.” This isn’t denial. It’s grief.

When you’ve built your worth and your pleasure on coding better than others, and a machine produces acceptable code en masse, there’s something to bury. It’s not irrational. It’s human.

Why the craft has never been more alive

I’ve been coding for nearly thirty years, and I’ve never found it more useful than in 2026.

Because what AI destroys (producing code by hand, the programmer’s role) was never the heart of the craft. It was the visible part, not the essential one. The essential part has always been on the analyst’s side: understanding a problem before writing a single line, formulating the right intent, carving up a system so it stays understandable, judging whether a solution will hold up in six months, arbitrating between speed, debt, and real value.

None of that can be delegated to a model today. And AI doesn’t devalue these skills: it amplifies them. What it mostly weakens are those who had confused producing code with doing engineering. We confused them because, ever since the two jobs merged, you had to do one to do the other. That’s no longer the case.

I’ve always seen our craft as that of people who are of service because they spend their time learning. And from that point of view, AI is the best gift I’ve received in a long time: it lets me learn infinitely faster. A subject that would have taken me three weeks to clear on my own, I now explore in an afternoon.

What if the bubble bursts?

People sometimes ask me: “what if all this collapses, and you’ve forgotten how to code?” It’s a fear I take seriously. My answer isn’t to bet on the tool’s longevity: that would be falling back into the very mistake I’m calling out. It’s to keep the fundamentals alive. I’m doing algorithm exercises again, I work on data and ML. Not out of nostalgia, but because these are precisely the analyst’s muscles, the ones that will serve me whether AI stays or disappears.

The craft isn’t dying. It’s returning to its center.

Yes, there’s an identity crisis, and it’s legitimate. But it mostly comes from having confused our craft, for too long, with the act of writing code, with the programmer’s role, when we were also, and above all, analysts.

Our real craft has always been to learn, to understand, to structure, to make reliable, and to be of service. Seen that way, AI doesn’t kill the craft. It gives us back the analyst’s role, and hands the rest to the machine. It pushes us back toward our center.

The template ruler is still in my drawer. The tool is gone, the gesture is gone, but what we did with it (thinking through a system before building it) has never stopped being the work. The job has been dying for 40 years. And that’s precisely why it’s still here.