All articles

AI won't replace you. Running out of value will

The hiring test that matters: the person you would hand the company keys to for a week

Nobody hires out of charity

There is a lot of uncertainty among developers right now. Some are afraid AI will replace them. Some think years of learning just lost their value. Some are no longer sure what their role is. I wrote about it on LinkedIn recently, and this is the full argument that did not fit in a post.

The short answer: the reason a company picks a person has not changed. A company works with you as long as your work helps it grow, whether that means billing more, shipping a product, or taking one further. Nobody hires out of charity, and nobody renews a contract to be nice.

That sounds cynical. I read it as good news, because it is the one rule AI has not touched. If anything it weighs more now. When writing code gets cheaper, what a company really pays for is everything else.

A train on rails

From 2018 to 2023 I worked in consulting, at Deloitte, Doing and Minsait. I learned a lot there and I don't regret it. But for much of that time I felt held back, like a train on rails. The route was already drawn. My job was to reach the next station on time.

In 2023 I went freelance for one specific reason: choosing who I work with. Startups I can actually add value to, and that I click with. It is still the filter I use to say yes or no to a project.

The work didn't shrink. If anything it grew. What changed is that I could use everything I had built up: the studying, the projects, the habit of breaking a problem down. And I could work with each startup as if the project were mine. In practice that means pushing as hard as I can to give whoever called me the tools to level up, not just closing the tickets I'm handed.

Treating the project as mine

One example, no names. I joined a startup to work on its website and product, and I got to see the whole picture. What customers saw. What prospects saw. What an investor would see.

That turned into proposals. Concrete improvements meant to catch the attention of those three groups, not to make the code prettier. We sat down and studied them together, then I put my head down and built. There were a few nearly sleepless nights, I won't pretend otherwise.

We got the result we were after, and it gave the business a real boost. The site and product were redesigned from scratch, and the startup went on to raise nearly €2M. It's on my about page too.

I still work with them, and not only on code. Sometimes it's tech, sometimes business, sometimes internal productivity. Whatever is needed, with whatever I can do. And where possible, saving them money, which is a kind of value people tend to forget.

Who you would leave the company with for a week

Something we lose sight of: most of the time we are not saving the world, or anyone's life. We are bringing products and services to life, and trying to push them further.

That doesn't take the best programmer in the country. It doesn't take someone who knows twenty languages and a hundred libraries and has worked on every kind of project. Experience matters a lot, because it lets you spot a problem before it blows up. But it isn't what decides.

If I were hiring, this is what I'd look for. Someone who cares about the project. Who thinks through solutions and proposes improvements. Who doesn't stop at the first obstacle, and owns the job until it's done. That's the kind of person who adds value to a company, and it's one of the reasons I've always chosen to work with startups.

AI doesn't make these people less necessary. It makes them faster.

I used to be a React developer

Before AI I was already a fairly broad full-stack developer. React and Next.js, React Native with Expo SDK, Node.js, CI/CD pipelines. Deploys on Vercel, Cloudflare and Netlify, but also on-premise behind Nginx and Apache. I've seen a bit of everything. The rest stayed unexplored: between one commitment and the next, there was no time to go deep.

With AI I got to dig into things I barely knew:

• Tauri, to bring an app to the desktop;
• SwiftUI and Kotlin, for native iOS and Android;
• three.js, which stays a complex library even when you know what you want to draw;
• firmware for microcontrollers I had never touched.

How I got there matters more than the list. While I work through a problem with AI and the solutions to propose, libraries, languages and tools show up that I didn't know, or only knew by name. More options land on the table. I learn them by building with them, and that learning feeds itself: what I pick up on one product, I reuse on the next.

That is also the part AI isn't good at yet on its own: keeping a wide range of alternatives open when it's time to choose. It tends to hand you the most beaten path. Knowing there are three others, and which one will hold up better two years from now, is still your job. The more you've tried, the wider that range gets.

That's why I think we should see ourselves less as developers of one language and more as problem solvers. The language is the tool for the project at hand.

None of this means handing over my thinking. AI helps me build a lot more, but choosing what to build, how to build it, and guaranteeing quality, reliability and security is still on me. How I keep that in practice, with questions before code and reviews of my own PRs, is in Shipping fast is the easy part.

When the gap is closed

Being useful isn't a permanent state, and freelancing shows you that very directly.

I've worked with startups where I was needed up to a point. They needed a boost, or a technical gap filled, and after my time inside, the gap was gone. That's not a bad thing. It's often the whole reason to hire a freelancer: a short engagement where you leave the team everything it needs to keep going on its own.

With others the fit is so good that the collaboration runs for a long time, with stretches where I put in more hours and stretches where I put in fewer. And sometimes the budget runs out, or is about to, and the client has to make a call.

In all three cases, the question on the other side of the table is the same: what is this person bringing us? AI hasn't changed any of those endings. It has changed how much you get done before you reach them.

So when a short engagement is about to wrap up, I ask myself one thing. On Monday morning, can the team keep going without me? If the answer is yes, I did the job they called me for.