This is hard to write, partly because it feels like admitting something I’m not supposed to feel, and partly because I’m not entirely sure it’s true. But I keep noticing it, and ignoring it isn’t making it go away.
I’m losing interest in IT. And I think AI is why.
What I Actually Liked About This Work
Let me be honest about what kept me in this industry for as long as I’ve been in it. It wasn’t the salary, though that’s nice. It wasn’t the stability. It was the problem-solving. The particular satisfaction of staring at something that doesn’t work and methodically figuring out why, then making it work.
Debugging a cryptic kernel panic at 2am. Tracing a network issue through packet captures until you find the one malformed packet. Writing infrastructure code that you know is going to run reliably for years. Reading documentation nobody else read and finding the one obscure flag that makes everything click.
That was it. That was the thing.
I’m a hobbyist as much as a professional—running a homelab, building things I don’t strictly need, learning technologies because they’re interesting rather than because I need them for work. The curiosity was the fuel.
What Changed
AI coding assistants happened. And they’re genuinely good now. Not perfect, but good enough to handle the first 80% of almost anything—the boilerplate, the scaffolding, the pattern-matching against known solutions.
And here’s the thing I didn’t expect: I’m finding that 80% was most of what I enjoyed.
The debugging loop that used to feel like a puzzle now feels like watching someone else solve the puzzle while I supervise. I describe the problem, the AI produces a solution, I review it, it works or I iterate. The intellectual engagement is… muted. There’s still some problem-solving, but it’s shifted toward “is this AI output correct?” rather than “how do I figure this out myself?”
I used to read documentation. Actually read it—deep-diving into man pages, RFCs, Kubernetes API references, kernel docs. Understanding the system at a level below what I needed. Because understanding things felt good.
Now I ask the AI. And the AI answers correctly enough that the deep-dive feels like optional gold-plating.
The Skill I’m Worried About Losing
What I’m really worried about isn’t my current productivity—that’s arguably higher than ever. It’s that I’m not building the mental models I used to build. The ones that let you reason about new problems from first principles, make connections across domains, understand why a system behaves the way it does.
You build those models by struggling. By reading the thing you don’t understand until you understand it. By debugging something for three hours and having the moment of comprehension when it finally clicks.
If AI handles the struggle, do you still build the model? I’m genuinely not sure. And I don’t think enough people are asking this question.
The Craft Problem
There’s a version of every technical skill that’s craft—the difference between someone who can build something and someone who builds it well. Code that doesn’t just work but is clear, maintainable, elegant. Infrastructure that’s not just functional but thoughtfully structured.
Craft comes from care. From caring about the details even when nobody’s checking. From having an internal standard that isn’t “does it work?” but “is it good?”
When AI handles the implementation, the craft question becomes “is the AI’s output good?” rather than “is my work good?” And I notice myself caring less about the answer. Not because I’ve stopped valuing quality—I don’t think I have—but because there’s something psychologically different about curating AI output versus producing your own work.
This might be temporary. It might be a transitional phase while I figure out how to integrate AI into a workflow that still feels meaningful. But right now, it doesn’t feel great.
What It Might Mean
I want to be careful not to overclaim here. “AI is causing burnout” is a hot take that might be wrong.
Some alternative explanations:
- I’ve been doing this for a long time and some loss of novelty is natural
- The specific things I’m working on right now are less interesting than things I’ve worked on before
- AI is exposing that I was solving a lot of repetitive problems and calling it interesting when it wasn’t
- I’m romanticizing the pre-AI struggle
The last one is probably partly true. I didn’t actually enjoy every debugging session. Some of them were just tedious. AI removing tedium shouldn’t feel like a loss. But it kind of does.
What I Think Is Actually Happening
I think the honest answer is that AI has compressed the feedback loop between “I have a problem” and “the problem is solved” in a way that removes a specific kind of engagement. The kind that comes from sustained effort, from the gradual satisfaction of building understanding, from the flow state of deep work on a difficult problem.
The problems that remain interesting are the ones where AI doesn’t just give you the answer—the genuinely novel architecture problems, the multi-system debugging that requires understanding context AI doesn’t have, the judgment calls that require experience and values. Those still engage me.
But they’re a smaller fraction of the work than they used to be.
The Career Implications
I think a lot of people in IT are going to confront some version of this. Not AI taking their job—that’s a longer conversation—but AI changing what the job feels like in ways that weren’t anticipated.
The response, I think, is to deliberately move toward the parts of the work that AI doesn’t (yet) flatten. Architecture and system design. Security research. Novel problem domains. The work that requires deep context, institutional knowledge, and judgment built over years.
The technicians are going to struggle. The engineers who understand systems deeply are still valuable. The difference is who you’ve been training to be.
I’m not ready to leave this industry. But I’m paying attention to these feelings in a way I wasn’t two years ago. And I think that’s healthy.
At minimum, writing this down made me think harder about it than I have been. That counts for something.
