AI Burnout Is Real, and Nobody Wants to Say It Out Loud

Every week my feed is full of people showing what AI made faster. A feature in an afternoon. A prototype before lunch. Ten agents running in parallel. Almost nobody posts about how tired they are.

But talk to engineers in private and the story changes. They are exhausted. Not the old kind of exhausted, after a long release or a bad on-call week. Something slower and harder to name. A constant low hum of being behind, of switching, of never quite finishing.

I think that is AI burnout. And I think we need to start saying it out loud.

Why Nobody Talks About It

Admitting you are struggling with AI sounds like admitting you can’t keep up. And in this market, nobody wants to be the person who can’t keep up.

The tools are sold as relief. They are supposed to take the boring work away and make everything easier. So if you feel worse, the natural conclusion is that the problem is you. You are using it wrong. You haven’t found the right workflow yet. Everyone else seems fine.

Everyone else is not fine. Everyone else is posting the highlight reel. The demo that worked, never the six that didn’t. The feature that shipped in a day, never the evening spent untangling what the agent broke along the way.

There is also no clean story to tell. Burnout from overwork has a shape: too many hours, too much pressure. AI burnout often happens while the hours look normal and the output looks great. It is hard to complain about a day where you shipped more than ever. So people don’t.

The Context Switching Tax

The biggest change in my own days is not how much I build. It is how often I switch.

Write a prompt. Read the output. Spot something off. Correct it. Jump to another session that finished. Review a diff in code I didn’t write. Answer a question from a third one. Go back to the first and try to remember what I was even checking.

Every one of those switches is small. Together they are brutal. The brain is not built to hold five half-finished problems at once and rotate between them every few minutes. Flow used to be the reward of engineering: an hour where the problem and you were the only things in the room. That hour is getting rarer.

I wrote earlier this year about how the new burnout comes from watching, not building. Supervising a machine that never stops is its own kind of tiring. Context switching is what makes it worse. You are not just watching one system. You are watching several, and each one pulls your attention whenever it wants.

The result is a strange feeling at the end of the day. You were busy every minute. A lot got done. And nothing feels finished, because you never stayed with one thing long enough to feel it land.

The Treadmill of Keeping Up

Then there is the second load: the pace of change itself.

A new model every few weeks. A new coding tool, a new agent framework, a new “this changes everything” thread. The workflow you set up three months ago is already called outdated by someone with a big following. The skill you spent a weekend learning gets absorbed into the next release.

For most of my career, staying current meant reading a bit, trying a new framework now and then, and going to a conference once a year. Knowledge had a long shelf life. Now it feels like it expires in months. And because the changes are real, you can’t simply ignore them. Some of them really do matter.

So you try to follow everything. You read in the evening. You test the new tool on the weekend. You feel guilty when you don’t. Keeping up turns into a second job with no end date and no finish line, and the worst part is that falling behind starts to feel personal. Not “I haven’t tried that yet,” but “I am becoming obsolete.”

What It Does to Your Head

I am not a psychologist, and I won’t pretend to be one. But I can describe what I see in myself and in the engineers I talk to.

A constant background anxiety. Something new is always coming out. Someone is always further ahead. The feed never says “you are fine.”

Self-worth tied to output. When a machine can produce code faster than you, it is easy to start measuring yourself against it. That comparison is rigged, and it quietly eats at your confidence.

Trouble switching off. Agents can run while you sleep. So part of your brain stays on duty. You check a session from the couch. You start one more task before bed because it will be done by morning.

Identity questions. For experienced engineers especially. If writing code was the craft you loved and the thing you were known for, what are you now? A reviewer? A prompt writer? I touched on this in AI Made Writing Code Easier. It Made Being an Engineer Harder. The question hasn’t gone away. It has only become more common.

None of these show up in a sprint report. That is exactly why they get ignored.

What Actually Helps

I don’t have a cure. I do have a few things that have made a real difference for me and for people I work with.

• Pick your tools and stop shopping. Choose a small set that works, and ignore new releases for a while. You can evaluate once a quarter instead of every week. Most of the “game changers” will still be there, or will be forgotten.
• Run fewer things at once. Two agent sessions you actually understand beat five you skim. Parallelism is only a win if your attention can keep up with it.
• Keep one block of the day without AI. Read code properly. Write a design note. Think about a problem nobody has framed yet. It reminds your brain what deep work feels like.
• Set a hard stop for agents. If nothing runs after a certain hour, there is nothing to check. Your evenings come back.
• Measure outcomes, not output. Did the thing that mattered get better? That is the question. Lines generated and PRs merged are not.
• Talk about it with your team. Say “this week was too much” in a retro. You will be surprised how many people were waiting for someone else to say it first.
• Get real help when it is more than tiredness. If you can’t sleep, can’t switch off, or stop enjoying things outside of work, talk to a professional. That is not weakness. It is maintenance.

And if you lead a team, this one is on you more than anyone. Your engineers will copy your pace. If you answer agent sessions at midnight and post about every new tool, they will assume that is the standard. Show them a calmer one.

Saying It Out Loud

AI is changing our work, and a lot of that change is good. I use these tools every day and I am not going back. But tools that make us faster don’t automatically make us well. Sometimes they do the opposite, and pretending otherwise only leaves people to struggle alone.

So here is my small contribution: I get tired from this too. The switching, the pace, the feeling of always being one release behind. You are not doing it wrong. It is just hard.

How are you actually doing with all of this? Not the highlight reel. The real answer.

If this resonated with you, I would love to hear your perspective. Drop me a message or connect with me on Twitter (X) or LinkedIn. I write regularly about the intersection of AI, software engineering, and leadership at ivanturkovic.com.