“I don’t know why this week became the tipping point, but nearly every software engineer I’ve talked to is experiencing some degree of mental health crisis…” Many people assuming I meant job loss anxiety but that’s just one presentation. I’m seeing near-manic episodes triggered by watching software shift from scarce to abundant. Compulsive behaviors around agent usage. Dissociative awe at the temporal compression of change. It’s not fear necessarily just the cognitive overload from living in an inflection point.

— Tom Dale¹

That appeared on my timeline this week and I read it twice. Then I sat with my coffee getting cold and read it a third time, because Tom had named something I’d been circling for months without quite landing on.

It’s not that I hadn’t noticed my own disquiet. I’d written about nostalgia for writing code, about the strange grief of watching a craft you’d spent years developing get partially automated. But I’d been framing it as a productivity question, an ownership question. The quote reframed it as a health question, and that felt more honest.

What this actually feels like #

What I notice in myself isn’t primarily job anxiety, though there’s plenty of that around too. The 38% of employed adults who worry AI will make their job duties obsolete² are worried about a recognisable thing: income, relevance, being pushed out. In the same APA survey, 64% of those worried about AI reported feeling tense or stressed during the workday, compared to 38% among those without that worry. Those are real and measurable pressures, with a shape you can reason about.

What sits underneath that, for me, is stranger. It’s the cognitive texture of watching an entire field change faster than you can process it: not a single disruption you adjust to and move on from, but a continuous destabilisation where every tool you learn is superseded before it’s fully absorbed, and every position you form about how to work gets complicated by something that shipped while you were forming it.

I went looking to see if there’s research on this particular experience and mostly found gaps. The psychological literature has frameworks for technology-induced stress, for burnout, for cognitive overload. What it doesn’t have yet is a vocabulary for what the quote describes: near-manic episodes, dissociative awe, compulsive agent usage. That language is ahead of the science, which is its own strange sensation.

The productivity paradox makes it worse #

There’s a particular cruelty in one finding I keep coming back to. Experienced developers who expected AI assistance to reduce task completion time by around 24% actually took about 19% longer with AI than without it.³ The tools feel powerful. The numbers don’t confirm the feeling. And rather than prompting scepticism about the tools, this tends to prompt self-criticism: I must not be using them right, I must not be prompting correctly, I haven’t figured out the workflow yet.

The tools are new and strange enough that their failure modes feel like your failure modes. So you spend more time with them trying to close a gap that may not actually close, while the cognitive load of learning and re-learning and doubting your own assessment keeps rising. Rising cognitive load contributes to fatigue, errors, and reduced capacity to think creatively,⁴ which means the thing you’re using to keep up may be part of what’s making it harder to keep up. The self-criticism, the sense of being behind despite using the accelerant, the sustained pressure of continuous re-calibration: these aren’t separate frustrations. They compound into the kind of psychological strain that affects how clearly you can think, which feeds the self-criticism, which sends you back to the tools.

I don’t say this to dismiss AI tools, which I’ve found genuinely useful. But I’ve noticed in myself a kind of compulsive checking: has something changed, has a new model dropped, is what I learned about this workflow last month already outdated? It doesn’t feel like curiosity. It feels like vigilance.

Reaching for something older #

I’ve been re-reading Marcus Aurelius lately. Not as philosophy homework, more as a kind of instinct, the way you reach for a particular book when you can feel yourself needing something it has.

Aurelius wrote the Meditations during the Antonine Plague, which killed somewhere between five and ten million people across the Roman Empire. He was also fighting wars on multiple fronts, managing an empire under conditions of continuous crisis, and by his own private account (the Meditations were never meant to be published) struggling, repeatedly, to hold himself together. What strikes me reading him now is not the famous aphorisms but the repetition. He writes the same lessons to himself again and again, in slightly different language, as if he can’t quite make them stick. The equanimity people associate with Stoicism was not a settled state he’d achieved. It was a daily practice he required because conditions kept demanding it.

That feels true to where I am. Not collapsed, not in crisis, but in need of something to practise against the noise.

The distinction Aurelius returns to most, and that I keep finding useful, is between what is within our power and what is not. What you can govern is your own response, your own judgment, how you meet what happens. What you cannot govern is the event itself: the rate of change in this industry, what ships next month, whether your current skills retain their value. Seneca makes an adjacent observation that I also keep coming back to: most people don’t experience time as stolen, they give it away, because distraction feels like activity. I recognise that in myself. The compulsive checking, the constant reorientation to new tools, the sense that stopping to think is falling behind.

I find myself genuinely uncertain whether this framework helps or whether I’m using it to avoid the discomfort of actually sitting with how strange this moment is. But Aurelius didn’t seem convinced it was working either, which may be the point.

The mental health data underneath it all #

In a 2023 survey of 500 software professionals, 50% reported having done therapy at some point, and 21% were currently in therapy.⁵ I don’t know how those numbers compare to the general population but they don’t feel low, and they predate the current pace of change. Whatever the psychological burden of software work was before the last two years of acceleration, it seems reasonable to assume the current moment hasn’t lightened it.

The phrase in the opening quote that keeps pulling at me is “temporal compression of change.” A precise description of something I don’t have a better name for: change happens fast enough that there’s no stable present to get your bearings in, and you’re always adjusting to something that’s already been superseded.

I’m not sure the Stoics fully anticipated this, though they had their own versions of it. Aurelius watching the Roman world his predecessors had built begin to fray. Seneca writing under Nero, never quite sure whether the next summons would be a conversation or an order to kill himself (it was eventually the latter). The philosophical tradition they developed was built for genuine instability, not abstract instability. That’s part of why it keeps feeling relevant.

What I don’t know #

The honest version of this is that I’m writing it partly to find out what I think. There’s something in me that wants to resolve this into a framework, a set of practices, a conclusion about how to navigate this well. That impulse is probably worth resisting.

Aurelius wrote in one of the later books of the Meditations: “You have power over your mind, not outside events.” He wrote it to himself. He probably wrote it because he kept needing the reminder. I find that more useful than the polished version of the same idea, the one that sounds like a resolution. The real thing is messier. It’s a practice you take up again each morning not because you’ve figured it out, but because the alternative is worse.

I’m still figuring this out. I’m not sure that changes.