Every conversation about AGI eventually arrives at the same uneasy question: if machines can do the valuable work, how should people spend their time? It sounds like one question. It's actually two, and most of the dread comes from letting them blur together.
The first is economic. If AI automates most of what we currently get paid for, how do people get income, structure, and status? That's a real and hard problem, and the transition could be turbulent even if the destination is good. But it's a question about how society organizes itself, and nobody - including me - actually knows what form it takes.
The second question is the one that keeps people up at night: when work is no longer necessary, what makes a life worth living? And here's the thing I've come to believe, the more I sit with it. AGI doesn't change that answer nearly as much as it seems to. What AGI threatens is the economic necessity of work. It doesn't threaten the human capacity for meaning. Those two things were always more separate than we pretend.
Keynes Already Ran This Experiment in His Head
In 1930, John Maynard Keynes wrote a short essay called "Economic Possibilities for our Grandchildren." He predicted that by around 2030, productivity would have risen so much that people would work about 15 hours a week. He was wrong about the hours - UK working time fell by roughly a quarter over the following 80 years, not the two-thirds he forecast, and economists like Nicholas Crafts have shown most of the freed-up time went into longer retirements and lifespans rather than shorter weeks.
But the interesting part isn't the miss. It's what Keynes thought we'd face if the prediction came true. He called it humanity's "permanent problem" - how to use freedom from economic necessity "to live wisely and agreeably and well." And he wasn't optimistic about the transition. He predicted a kind of collective "nervous breakdown," pointing to the wealthy of his own era, who had already solved their economic problem and largely failed to do anything worthwhile with the freedom.
That's the honest worry. Humans are historically bad at unstructured abundance. Lottery winners, early retirees, idle aristocracies - freed time becomes a good life only when people use it actively, and reliably curdles when it's just handed over. So "should" is the right word in the question. None of this happens automatically.
The Premise Is Already Arriving
You don't have to believe in science-fiction AGI for this to be live. The IMF estimates about 40% of jobs globally are exposed to AI, rising to roughly 60% in advanced economies. Goldman Sachs puts about two-thirds of U.S. occupations as at least partially exposed, with around a quarter of all work potentially automatable when weighted by how many people do it. McKinsey's 2025 estimate is that current technology could, in theory, automate over half of today's work activities.
Notice what's interesting in the more careful versions of this research. PwC's 2026 analysis of over a billion job postings found that the tasks being added to AI-exposed roles are about 2.5 times more likely to lean on empathy, judgment, and creativity. As AI absorbs the routine, the distinctly human parts of work don't vanish - they become the whole job. That's a small preview of the larger pattern.
"But the Machine Does It Better"
Here's the deepest version of the doubt: if AGI does everything better than I can, doesn't my contribution become hollow? Why master anything? Why try?
I think this objection mostly dissolves, and the way it dissolves is worth sitting with, because we've already run the experiment.
In 1997, Deep Blue beat Garry Kasparov. By the late 2010s, chess engines were so far beyond human ability that the world's best players study from them, not the other way around. The machine didn't just beat us - it lapped us, permanently, decades ago.
So chess should be dead, right? Why play a game a $30 phone plays better than any human alive?
Chess is more popular than at any point in its 1,500-year history. Chess.com crossed 100 million members at the end of 2022, hit 200 million by April 2025, and passed 250 million in early 2026 - roughly 12 million new players a year. More than 6 billion games were played on the platform in 2024 alone. Total, machine-crushing superiority did nothing to the human appetite for the game. If anything, the boom accelerated after the machines won.
The same logic is everywhere once you look. Cars are faster than any runner, and millions of people run marathons. Restaurants exist, and people still cook dinner for the people they love. The value of these things was never in producing the optimal output. It was in the doing, and in who you did it with. Strip away the instrumental justification and you don't kill the activity - you purify it. It becomes honestly about the experience.
What Survives Is the Non-Fungible
So which activities survive contact with AGI? The ones where the value is non-fungible - where it has to be you, or where the worth lives in the effort and the relationship rather than the result.
Your kid doesn't want a more optimized parent. Your friends want you at the run on Tuesday, not a superior substitute who shows up in your place. Embodied skill you built over years, a body moved through real effort, relationships accumulated by showing up again and again with the same people - an AI doing any of it "better" somewhere else doesn't touch a single bit of it, because being the best available method was never the point.
This is why I don't think the post-AGI menu of meaning is new. It's the same list humans have always pointed at when they weren't being forced to earn a living: make things, learn hard things badly and then less badly, take care of people, move your body, build a few relationships deep enough to be irreplaceable, attend to something larger than yourself. The shift isn't what counts. It's that these stop being what you do after the real work and become the real work.
The Real Danger Isn't Idleness
If there's a genuine risk in all this, I don't think it's mass despair or empty days. I think it's something quieter and already here: the slide toward frictionless passive consumption - endless, beautifully engineered stimulation that is pleasant and nourishes nothing.
Look at how we already spend the free time we have. According to the U.S. Bureau of Labor Statistics' American Time Use Survey, the average American has about five hours of leisure a day - and spends more than half of it, around 2.6 hours, watching TV. The amount left for actually socializing with other people is about 35 minutes. And that's shrinking: in 2014, 38% of people socialized on a given day; by 2024 it was down to 30%.
The cost shows up in the loneliness numbers, which are genuinely alarming. The U.S. Surgeon General's 2023 advisory found about half of American adults experiencing loneliness, with in-person time spent with friends falling from roughly 60 minutes a day two decades ago to about 20 minutes by 2020. For people aged 15 to 24, time with friends dropped by about 70%. The health toll is comparable to smoking 15 cigarettes a day.
We are not, it turns out, idle. We are busy - busy consuming. We already have abundant leisure and we are already pouring it into the most frictionless option available. AGI doesn't introduce this temptation. It sharpens it, by making the engineered stimulation even better and the path of least resistance even smoother.
Aristotle drew a distinction that's old but exactly right for this moment: leisure as mere recovery from work, versus leisure as activity pursued for its own sake, which he considered the highest form of human life. A world of automated work removes the need for the first kind and clears enormous space for the second - but only if we actually reach for it.
The Answer Is Participation
So, how should people spend their time if we reach AGI?
Roughly the way humans have always been at their best: doing real things, with real people, repeatedly. Not spectating - participating. Not consuming experiences - having them, with the same faces, often enough that bonds and skills and shared history actually accumulate. The recurrence is the mechanism. You don't become someone's friend by attending one great event; you become their friend by showing up to the ordinary one for the tenth time.
This is the part I've staked my working life on, so take it with that grain of salt. But I'd believe it even if I weren't building toward it. An age that automates cognitive work doesn't make showing up, in person, with the same people, doing something together, less valuable. It makes it the most valuable thing left - because it's the one thing that was never about output in the first place.
The machines will get better at almost everything. They will not get better at being you, in a room, with the people who came back to see you again. That was always the point. We're about to find out whether, freed from necessity, we choose it.
← Back to blog