The Work AI Leaves Behind

I was seventeen when I got laid off for the first time.

That sentence feels ridiculous to write. At seventeen, I should have been worried about prom, acne, and whether my car was going to start. Instead, I was being pulled into a manager’s office in a building that looked like it had been designed by people who were afraid of dust. The floors were marble. The walls were clean. The stainless steel looked like someone had just finished polishing it with a toothbrush. I worked long shifts there, partly because I liked the nighttime quiet. After everyone left, the building settled into this sterile little hum. Air conditioners. The occasional printer. My own footsteps sounding more important than they were.

Then I was whisked into an office and told the company was changing direction. My manager’s face was red. He looked angry, but not at me. He told me he was upset because, whether I moved to IT or took the layoff, he was losing a valuable employee. I wish I could tell you I received this with maturity and grace, but I was seventeen, so of course I acted like a wounded prince. I could not believe this was happening to me. After all I had done for the company. All those late nights. All that effort. All that marble I had heroically walked across.

They gave me a choice. I could move from R&D into IT, or I could take a severance package and leave. Before I decided, they gave me a tour of the IT department. It was like stepping out of a hospital lobby and into a daycare for adults. Every wall was a different bright color. Yellow. Orange. Red. Blue. The manager was as chipper as a children’s TV host on picture day. She explained that the colors were chosen to inspire different emotions. Red walls were for people who needed to work faster. Blue walls were for people who needed to calm down customers. The marble and stainless steel were gone. Now there was carpet and couches and emotional paint.

I chose severance.

The Part I Missed

At the time, I thought the company had done something to me. That was not completely wrong. Losing your job is personal when it is your paycheck, your routine, and your sense of being needed. But it was also not the whole truth. The company was pivoting. The work I had been doing was not as central to the future they were building. They did not say I was useless. They said the old shape of my usefulness did not fit the new shape of the company.

That is much easier to type than it was to live. I did not hear, “The company is changing direction.” I heard, “You are replaceable.” I did not hear, “Here is a new path if you want it.” I heard, “Your old path is gone, and also the new path has orange walls.” In fairness to my younger self, the orange walls were a lot. But I was so focused on the insult that I almost missed the lesson.

The lesson was not that loyalty is fake or companies are heartless. Sometimes those things are true, but that is too easy. The harder lesson was that work moves. A title can disappear while the person’s value remains. A department can shrink while a new department grows on the other side of the building, wearing bright colors and asking you to become someone slightly different than you were yesterday.

The Programmer Version

I think about that old office when programmers talk about AI taking jobs. I hear the same fear I felt in that manager’s office. Something is changing. The work that made me useful may not be the work that makes me useful next year. Someone is walking me down a hallway toward a room I do not want to sit in, and they are telling me the walls are blue for my own emotional benefit.

AI is replacing real programming work. I do not want to soften that until it turns into a lie. It can generate code, explain code, write tests, translate from one framework to another, summarize unfamiliar files, and produce the kind of boilerplate that used to consume an afternoon. Some programming jobs are built mostly out of that work. Those jobs are going to feel pressure. Some of them will go away.

I am not going to speak for writers, artists, designers, accountants, lawyers, or anyone else whose field is being dragged into the AI conversation. I am a programmer. That is the door I can hear the knocking on. And for programmers, the choice is becoming hard to avoid. Either learn the new shape of the work, or go find somewhere that still needs the old shape.

The Work Moves

Later in my career, I saw this pattern again in doctor’s offices. I would come in, look at a workflow, and find some poor soul moving information from one place to another all day. Click. Type. Print. Scan. Call someone because the scanned thing did not match the typed thing. It was the kind of work that makes computers look like a mercy.

So we would put part of the workflow into a system. The office got time back. Accuracy went up. The person who had been fighting the same forms every day got to stop fighting quite so many of them. But the headcount usually did not drop. The job changed. Often, the same person became the one who managed the system, checked the exceptions, and knew when the computer had done exactly what we told it to do instead of what we meant.

Startups had their own version of this. A new tool would show up, and everyone would get excited because it looked like it could replace one or two programmers. I was usually excited too, because I am exactly the kind of person who sees a new technology and thinks, “I can make this do my chores.” Then we would wire it into the company and discover a new job hiding inside the tool. Someone had to understand it. Someone had to know what good output looked like. Someone had to explain to a customer that the robot lied to us, but in fairness, we gave it a confusing spreadsheet.

The Numbers Are Messy

The research does not give us a neat bedtime story. The International Labour Organization looked at generative AI in 2023 and found that most jobs were only partly exposed to automation. Their read was that generative AI was more likely to complement most jobs than replace them outright, though clerical work had especially high exposure. That is not a programmer-specific forecast, but the shape matters. The machine can take a big bite out of a task. The whole job is harder to swallow.

The World Economic Forum’s 2025 jobs report is more dramatic, but still not as simple as the panic version. It projected 170 million jobs created and 92 million displaced by 2030, for a net gain of 78 million jobs. I do not treat that like prophecy. It is a projection, and projections are just guesses wearing a nicer shirt. But it is useful because it points toward disruption, not simple disappearance.

Daron Acemoglu and Pascual Restrepo have a helpful frame for this. They talk about automation having a displacement effect, a productivity effect, and the possibility of creating new tasks. In normal-person words, some work gets taken away, some work gets cheaper, and some work appears because the new machine exists. All three can be true at the same time. That is why this conversation is so hard. You cannot pay rent with net job growth by 2030 if your job is the one that disappeared on Tuesday.

The Old Pipes

When I was a young programmer, I thought old programming languages died like old phones. One day everyone had one, then a few people had one, then it ended up in a drawer next to a mystery cable nobody could identify. Then I found out some Fortran programmers were making very good money, and my tidy little theory fell apart.

Fortran was old, but it was not dead. There were scientific and engineering systems that still depended on it, and there were fewer people who could work on those systems without creating expensive smoke. A 2021 paper on the Fortran developer community described modern Fortran as one of the dominant languages for compute-intensive scientific and engineering applications. That surprised me because I had confused old with irrelevant, which is a common mistake among young programmers and people shopping for furniture.

Sometimes old work becomes rare work. Sometimes rare work becomes expensive work. Sometimes the valuable person is not the one who knows the newest tool first, but the one who can walk between the old room and the new room without pretending the old room never mattered.

The Forecast

If you are close to retirement, I would probably dig in. That is not career advice. That is just me being practical. There will still be old systems, old tools, old companies, and old processes that need someone who knows how they work. Some places will adopt AI slowly. Some will adopt it badly. Some will keep paying people to maintain the thing everyone else forgot how to touch.

If you are in your prime, I think the better play is to become the bridge. Teach people the transition while you make it yourself. Learn the tools deeply enough to tell the difference between a party trick and a workflow. Help your company move from the old room to the new one without pretending the old room was stupid. That may be the most valuable programming work for a while.

If you are just starting, transition now. I would not spend ten years becoming excellent at the parts of programming AI is already good at. Learn them enough to understand the craft, because fundamentals still matter. But do not build your whole identity around typing boilerplate faster than the next person. The machine is rude, tireless, and does not need snacks.

The Choice

AI will come knocking on many doors. That is an uncomfortable truth, because a lot of us are holding our breath and hoping it is not ours. For programmers, I think it is already on the porch. Maybe it is not kicking the door in today, but it is not across town either.

If it is your door, you have to look past the prejudice and the ego. I say that as someone who did not do this well the first time work moved under my feet. It is easy to get offended by the new room. It is easy to mock the colors, the couches, the language, the people who seem too excited about the thing that is making you feel replaceable.

But once you are done being mad, the choice is still there. You can learn the new thing, or you can keep doing the old thing somewhere else. Both are real choices. Both may be reasonable. The only option that will age badly is pretending no choice has to be made.

Look past the wall color.

Sources woven into this entry: International Labour Organization on generative AI and jobs, World Economic Forum Future of Jobs Report 2025, Acemoglu and Restrepo on AI, automation, and work, and Toward Modern Fortran Tooling and a Thriving Developer Community.