Technostress

This week brought back Fable. It also brought back something I hadn't seen in months. The session limit notification. That rather annoying message telling me I've used up all my AI turns for now. I have a few AI subscriptions, so I could take the work somewhere else outside of Claude Code and carry on. I don't. Instead, this week I felt myself taking a pause. The screen tells me I can go again at 6:40pm, and part of me is already counting down to it, lining up the new prompts. Some people call these tools the new (gateway) drugs. I can see why. The feeling of one more hit is so tempting.

There's a name for this, and more and more academic research is coming out about it. A recent review of technostress in the era of generative AI describes the same loop I, and many others, are in: reaching for the tool on reflex, then using up just as much energy watching it to see whether it actually helped versus doing it without.

Somewhere in the last year, staying fully productive stopped feeling optional. Not just for me. Doing anything by hand started to look like a small failure of imagination. We started optimising everything. We built tools to do the work, then tools to manage the tools. Then it turned fully personal and automagic; auto mode arrived and it was, genuinely, magical. Hand off a task and step away. So on an average day, I hand off four tasks at once. Customer work in one window, a side project in another, something half-formed in the third and fourth. For a while it felt like pure leverage.

Then AI fatigue kicked in. It didn't come from the work. It came from managing my own attention across three or four agents at once, none of which ever tell me to stop. AI is built to keep pushing you to do more. The researchers from the article above draw a line between using these tools as a scaffold and using them as a replacement. I know which side I and others are drifting towards. Nowadays, I reach for a model before I've finished forming the thought, and feel uneasy when nothing is running in the background.

Something odd happens when the Claude Code limit finally hits and I'm forced to stop. This week when I was in this moment of pause, the work got better. I think a problem through myself instead of handing it off. The focus I thought I'd optimised away comes back, because I'm not outsourcing the thinking any more. What felt like productivity was often just the habit that I formed. Four agents running is not the same as four problems solved.

None of this makes me anti-AI. Please don't think that. These tools are extraordinary and I'm not going back (is this even possible?). I just want to be honest about the side effects, because we don't talk about them enough.