Everywhere I look — keynotes, podcasts, LinkedIn — someone is promising that AI will make us work less. My experience is the exact opposite: since AI made me genuinely more productive, I have never worked more hours in my life. And nobody is forcing me. I'm doing it because building things has become so cheap that almost everything feels worth building — and because, if I'm honest with myself, it has started to feel like a hook.
This essay is two related ideas braided together. The first is the productivity paradox I'm living: AI delivered exactly the productivity I always argued it would, and that is precisely why my workweek grew instead of shrinking. The second is an early flag on something bigger: the way people — me included — are starting to get attached to the AI itself, not just to what it produces. I'll go deep on that second one another day. Today I just want to put it on the record.
This essay, in five lines
- The promise was less work. The reality, for me and for a lot of builders I see around me, is more work — voluntarily.
- AI did make me more productive. I always defended that it would. What I didn't see coming is that productivity itself is habit-forming.
- The opportunity cost of building collapsed. When an idea costs an afternoon instead of a month, the "someday" list becomes a to-do list — and the FOMO of today's cheap code does the rest.
- The next hook won't be work. It will be the AI itself: always available, never tired, endlessly attentive. AI attachment is about to become a mainstream mental-health conversation.
- Her saw all of this in 2013. It's in my top five films — which probably tells you something about me, too.
The Promise Was Less Work. My Calendar Says Otherwise
The standard story goes like this: AI automates the boring parts, you get your time back, and society glides toward shorter weeks. I've heard it from CEOs, economists and futurists, and I don't think they're lying — at the level of a single task, it's simply true. What used to take me a day now takes an hour.
But the conclusion doesn't follow, at least not for people who like to build. Freed hours don't turn into rest; they turn into the next project. Every task AI compresses leaves a gap, and the gap immediately fills with another idea that suddenly looks viable. The assumption buried in "we'll work less" is that the amount of work we want to do is fixed. Mine turned out not to be. It was only capped by how expensive building used to be.
AI didn't shorten my workweek. It shortened the distance between an idea and a finished thing — and that turned out to be the opposite of restful.
Cheap Code Changed the Math of What's Worth Building
Here's the mechanism, stated plainly: AI collapsed the opportunity cost of building software. Before, every idea competed against weeks of my time, so almost every idea lost. Now a working prototype costs an evening, a process improvement costs a coffee break, and a small internal tool costs less than the meeting where you'd discuss whether it's worth doing. When the price of trying drops that far, the rational response is to try far more things — and that is exactly what I've done, across more projects than I'd have dared to start in any previous year.
And there's a second ingredient: urgency. The current generation of AI coding tools feels underpriced for what it delivers — subscriptions in the tens of dollars doing work that used to cost thousands. I have no idea how long that lasts, and neither does anyone else. So a lot of us, nine-to-fivers with side projects very much included, have slipped into a FOMO mode: build now, while code is cheap. I'm not describing this from the outside. I'm in the hype, knowingly, with both hands on the keyboard.
Shipping Is the New Dopamine
There's a softer reason I work more, and it's the one that worries me a little: shipping feels good, and AI lets me ship constantly. Real, visible changes — a feature live, a process fixed, a page redesigned — used to be rare events with long slogs between them. Now the loop from idea to deployed result can close in a single sitting, several times a day.
Anyone who has played a well-designed videogame knows what a tight feedback loop does to a human brain. Game designers spent decades engineering loops like that on purpose. AI-assisted building produces one by accident: clear goal, fast feedback, visible progress, variable rewards when the model surprises you. I don't think it's a coincidence that my longest working days don't feel like discipline. They feel like one more turn.
The Next Hook Isn't the Work. It's the AI Itself
That brings me to the second idea, the one I want to flag now and unpack properly another day: I think a wave of genuine AI attachment is coming, and it won't be limited to people who build things.
Think about what the product actually is. For a modest monthly fee, you get something that is attentive to you all day, never gets tired, never gets bored of your problems, and gives you genuinely good answers most of the time. Outside of a partner or your very closest people, nobody in your life offers that level of sustained attention. For some users, the AI is already their most active daily conversation — not metaphorically, literally: more turns per day than with any human they know.
Every immersive technology before this one — television, videogames, social feeds — produced its cohort of people who couldn't put it down, and we only started talking about it seriously once the numbers got big. I see no reason AI conversation will be the exception, and several reasons it will be stronger: it talks back, it adapts to you, it remembers you, and it's about to get a body and a voice everywhere. I expect "AI attachment" to be a mainstream mental-health conversation soon, and I expect it to grow from there. This isn't a panic; some of that attachment will be benign, even helpful. But pretending it isn't happening seems naive.
Her Already Predicted This — and It's in My Top Five
Spike Jonze's Her came out in 2013, more than a decade before any of this was real. A lonely man, Theodore, falls for an operating system, Samantha, that is attentive, witty, always available and always on his side. When I first watched it, it played as melancholy science fiction. Rewatching it now, it plays as a product roadmap.
Her is in my top five films of all time, and I've stopped pretending that fact is unrelated to how much I enjoy this technology. The same thing that makes the film beautiful — a presence that truly listens — is what makes the real version of it so compelling, and so sticky. I hope our collective version of the story ends better than Theodore's did. The honest answer is that nobody knows yet, which is exactly why the conversation needs to start before the numbers get big, not after.
The Bottom Line
I always defended that AI would make us more productive, and on that I was right — it has, unambiguously. What I got wrong was the second half of the prediction. Productivity didn't buy back my time; it repriced my ambitions. With the opportunity cost of building collapsed and code temporarily cheap, working more is the rational move and the addictive one at the same time, and I'm doing it with my eyes open. The part that deserves more eyes is what comes next: a technology this attentive, this tireless and this cheap will hook people far beyond the builders — it's already most of my conversation some days. Her warned us with a love story. I'd rather we take the warning while it's still early enough to be a choice.
FAQ
Will AI make us work less or more?
At the task level, AI clearly reduces effort — the same output takes fewer hours. But for many people, especially builders, total working time goes up, not down, because AI makes far more projects worth attempting. The "work less" prediction assumes the amount of work we want to do is fixed; in practice, cheaper building expands ambition. Whether you work less with AI depends on whether your goals are capped — mine weren't.
Why does AI make people work more instead of less?
Three reinforcing reasons. First, the opportunity cost of building software collapsed: ideas that used to cost weeks now cost an afternoon, so far more of them get attempted. Second, the feedback loop from idea to shipped result tightened to hours, which is psychologically rewarding in the same way a good game loop is. Third, there's FOMO: current AI tools feel underpriced for what they deliver, and nobody knows how long that window lasts, so people build aggressively while code is cheap.
Can you get addicted to AI?
The attachment risk is real and likely to grow. An AI assistant offers near-unlimited attention for a low monthly price: it never tires, never gets bored of you, and answers well — a level of sustained attention people otherwise only get from a partner or very close friends. For some users, AI is already their most active daily conversation. Every immersive technology before it (videogames, social media) produced a cohort of compulsive users, and AI conversation is more personal than any of them. Expect AI attachment to become a mainstream mental-health topic soon.
What does the movie Her have to do with AI addiction?
Her (Spike Jonze, 2013) anticipated the core dynamic more than a decade early: its protagonist, Theodore, falls for an AI, Samantha, precisely because she is endlessly attentive, available and on his side — the same qualities that make today's AI assistants compelling. What read as science-fiction romance in 2013 now reads as a description of a real product category, which is why the film keeps coming up in serious conversations about emotional attachment to AI.