Nobody's making you work until 2am.

Nine things people carry

1. Too much coming in. A stream of output, diffs, and answers all day, faster than a person can take it in.
2. Higher expectations. The AI makes everything look fast, so more is expected, by others and by yourself.
3. Prompting, not learning. You prompt more and practise less, and your own skills feel like they're slipping.
4. The AI is better at parts of the job. It does some of your work, like writing code, better than you do, and that's hard to sit with.
5. Afraid to ship what you built. You built something with AI and now you're afraid to let real people use it.
6. Afraid to say “I don’t understand.” You rely on AI for answers too, and admitting you don't follow something feels risky.
7. What's left to solve. People love the ups and downs of a hard problem, and it can feel like soon there won't be any left.
8. What's real, what's next. Some days it's hard to say what's real, or what people are for.
9. Missing people. More hours with agents can mean fewer with humans.

These are what people describe, in posts and in conversations with us, not a clinical checklist. Most of them aren't about hours at all. They're about feeling capable, feeling that the work means something, and having people around.

The three patterns people keep describing

“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 external pressure required. Some developers describe working late because they enjoy the tools, even when they intended to finish earlier.

“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.

What a dashboard can honestly tell you

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.

What each feeling looks like in a week

• “Too much fun to stop” has a trace: the hour of your last check-in each night, and how much of the time your agents saved went straight back into new work launched after your planned stop.
• “Reviewer, not developer” has a trace: the balance between agent runtime and your own attention — approving, steering, fixing — per day and per agent, labeled as an estimate.
• “The mechanical weeks” don't have a clean number, and we won't invent one. What you get is your own baseline over weeks, so a week that felt flat can be compared against what your normal actually looks like.

What people say helped

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.

The short answers

• Why does working with AI feel so heavy? People describe nine things they carry: too much coming in; higher expectations; prompting instead of learning, so skills feel like they're slipping; the AI being better at parts of the job; fear of shipping what AI built; fear of saying "I don't understand"; wondering what's left for people to solve; questions about what's real and what's next; and missing people. If some of these sound like you, you're not the only one.
• Is it normal to feel like AI is making my skills worse? It's one of the most common things people describe: prompting more and practising less. What people say helps is keeping the hard thinking for yourself and doing some of the work with your own hands, on purpose, even when the agent could.
• How do I get peace of mind while my coding agent works? Put yourself first. Set the time your evening starts, step away while the agent runs instead of watching it, and spend the time back on something that's yours: a walk, a few pages, time with people. Teeras shows the minutes you have back (an estimate, always labelled) and suggests one thing sized to the window.
• Is AI burnout real? Enough developers describe it publicly — late nights that feel voluntary, reviewing instead of building, creative flatness — that the pattern deserves to be taken seriously. Whether AI work is extending your own day is measurable: when your agents worked, when you checked in, and whether those check-ins reached into the night.
• What are the signs of AI burnout? People describe three recurring signs: working late because the tools are engaging rather than because anything is due; feeling like a reviewer of AI output instead of a builder; and weeks where the work goes mechanical and creativity flattens. These are self-reports from public posts, not a clinical checklist.
• How is AI burnout different from regular burnout? People describe it arriving through enjoyment rather than pressure. Coding agents remove the natural stopping points — a build you would have waited for, a problem you would have slept on — so the day extends without anyone demanding it. The tiredness is described as real either way.
• Can Afterhours tell me if I'm burned out? No, and it won't pretend to. Afterhours shows when your coding agents worked, when you checked in, and whether check-ins reached into the night — counts and timestamps only, with human-attention figures always labeled as estimates. What the pattern means is your call to make.