AI burnout
That's the strange part. Developers keep describing a new kind of tired — not from deadlines, but from tools so engaging the day stops having an end. Here's what people report, in their own words, and how to see whether it's in your own week.
Three patterns keep showing up in public posts: late nights that feel voluntary, feeling like a reviewer instead of a builder, and weeks where the work goes mechanical. None of them shows up in token counts. They show up in when you check in — which is a thing you can actually measure, on your own machine, without uploading a single prompt.
“The people going hardest right now are burning out fast — working till 2AM not because they have to, but because they're having too much fun with the tools.”
from a public discussion of AI burnout, 2026
The first pattern is the odd one: it starts as enjoyment. Agents keep making progress, so there's always one more thing worth kicking off. No boss required. The old burnout had a villain; this one has a very good time, right up until it doesn't.
“When I hand over the problem-solving, not just the implementation, I stop feeling like a developer and start feeling like a reviewer.”
a developer using AI daily — at work and on a side project — posting publicly
The second pattern is about the role, not the hours. The same person wrote that keeping the hard thinking and delegating the execution made the work “feel like mine again.” The hours hadn't changed. What the hours were made of had.
“There are weeks where I'm so deep in prompts, outputs and iterations that my brain just stops engaging. Everything starts feeling mechanical. The creativity disappears.”
the same public thread on AI burnout
The third pattern is the quietest: nothing dramatic happens, the work just goes flat. It rarely arrives alone — it tends to follow weeks of the first two.
Not whether you're burned out. No app can diagnose that, and you should distrust one that claims to.
What a dashboard can show is the shape of your week: when your coding agents worked, when you checked in on them, and whether those check-ins reached into the night. Human-attention numbers are always labeled as estimates, because that's what they are. Whether your picture reads as “having fun” or “can't stop” is a call only you can make — the point is that you get to make it looking at your actual week, instead of at 2am with the terminal open.
The same posts that describe the problem usually end with what helped: keeping the hard thinking human and delegating the execution; one day a week fully offline from the tools; a stop time that actually holds.
Afterhours doesn't prescribe any of these. It gives you the before-and-after: pick one experiment, run it for a week, and see whether your evenings actually change — or whether they don't, which is just as worth knowing.
One command, on your machine. Counts and timestamps sync; prompts, code, output, and file paths never leave your computer.
npx @teeras/afterhours setup