Jevons Paradox: A personal perspective
What if the most game-optimal play in the new system is actually to become relentlessly, unapologetically you?
“I'm dying. Of overwork.”
When I read those sentences by Kache, I thought of all the late nights I spent creating just yet another mini app, a research deep dive, a short story. I indulge with exhilaration but also feel the creeping exhaustion, a good kind, almost romantic and evocative of the Renaissance. The hunger to unravel the world's secrets transformed sore wrists and weary eyes into marks of devotion, blissful in their ache.
This is the paradox of our time: the very tools designed to free us from labor are trapping us in an endless cycle of escalating work. As our productivity increases, our standards and expectations rise faster, creating a psychological Jevons Paradox that threatens to consume our humanity in the pursuit of ever greater output. We become victims of our own efficiency.
I sent the essay to some of my most prolific, creative friends, and they all resonated deeply. The indie hacker from Paris who has built four bootstrapped micro SaaS businesses with Loveable. Researchers in China taking apart viral closed-source products and launching open-source versions at record speed. The game designers from Prague who, with the integration of ThreeJS, can now build indie games with a Blender MCP, all playable within a browser, hosted on SpacetimeDB. The writer from LA that now, armed with Comfy and Midjourney, could storyboard their vision in high fidelity without pitching first to studios.
What happens when the consensual vision around how AI minimizes work time and maximizes leisure time is, in fact, wrong. And what happens in reality, when those who are already productive only work so much more? And as individuals become much more empowered, how does one stop oneself from working, that now the opportunity cost of not working becomes almost infinite?
Jevon’s Paradox: from coal to cognitive labor
If you work in tech or pay attention to AI news, you’ve heard about the concept of Jevons’ Paradox applied to the supply and demand of AI and computing. Investors justify the skyrocketing valuation of Nvidia stock with such a dynamic. Yes, there’s a flooding of supply, but more supply will only lead to consumption of more tokens, not less.
In 1865, when William Stanley Jevons noticed something peculiar about coal: making it more efficient to use led people to burn more, not less. This paradox has permeated every technological wave since — efficiency doesn’t tame our desires but whets them.
Compute and AI exhibit a strong version of this dynamic, as technical leaps in model architecture and infrastructure improvements rapidly drive down the cost per task. Potential demand for compute continues to expand as capabilities strengthen, with no saturation point in sight. As compute gets cheaper, possibilities multiply. Each new capability spawns a thousand use cases, and our consumption of resources blossoms with hunger.
And what happens when we apply this same logic to human labor? Our time, our energy, our creativity?
A century ago, in “Economic Possibilities for Our Grandchildren,” Keynes worried that productivity would grow so fast that our real problem would be leisure - what to do with ourselves once the economic problem got solved. We are now the grandchildren in his essay, and we are working more than ever. The shape of the work changed, and then there was more of it. Every hour saved gets reinvested, with interest
Coal engines once invented new appetites for energy; AI invents new appetites for labor — prompt engineer, RL engineer, AX designer, a title that did not exist when I started this essay. Whatever you gain gets eaten almost immediately, because the bar moves while you’re still working. If everyone can do 10x more, then 10x becomes the new baseline, and the edge you thought you’d won is gone before you know it.
This is Jevons Paradox, except it’s happening in your head. Coal got more efficient and we burned more of it, not less. Same here: the better the tools get at giving us back our hours, the more we expect from every hour, and we end up further behind our own expectations than when we started. That gap is the anxiety I hear from almost everyone I talk to — “I’m optimizing all of it, every part, all the time.”
A system is consistently in a state of entropy and only starts functioning when optimizing for set metrics, which dictate the system’s behavior beyond the participant’s control; so if you don’t decide what matters, the system decides for you. Scott Alexander describes the systematic involution as the Malthusian trap: competition pushes us to give up everything at the altar of productivity, until all the potential improvements get competed away, and there’s nothing left.
Measures of progress
Productivity was once a felt, corporeal experience: the joy of a full harvest with wheat in the barn, the satiating clink of a bin of metal parts. The main thing was to show up and keep working, labor measured with beads of sweat on the worker’s forehead. As those concrete tasks get supplanted by the omnipresence of bits on the screen, which have now evolved into their final form — superintelligence —productivity becomes no longer a concrete metric but a highly subjective experience, often characterized by a fluctuating sense of nihilism and hope. Am I getting somewhere, or am I just rearranging letters that have already been said? Is any of this real? Does this even matter? The competition used to be the person at the next desk over. Now it’s barely with anyone else, and half the time it isn’t even necessary, but it’s still there, and it’s with yourself, and what’s become hard is selecting a single door from an infinite corridor, and then no looking away.
I keep thinking about the months I spent in the Chinese tech scene as a college student for the summer. The infamous “996” work culture didn’t emerge from some conspicuous hustle porn, but rather from the existential necessity of retaining edge through endurance. For years, the West’s deepest worry was offshoring, a fear that high-skilled domestic labor would be replaced by tireless prodigies in Bangalore. Maybe that fear was misplaced. The work has, in fact, created brand-new categories that dissolve time zones and locations, and no matter who we are or where we are, we can contribute our knowledge through Mercor as little coefficients in a grand, seemingly endless matrix of next-token predictors.
Workers everywhere faced limitless opportunities, and accompanying that, pressure.
We have a rare chance to rethink how the world works from the ground up, but the best the industry can say about it is just a multiplier.
“Measuring what matters”
We've been handed a once-in-a-lifetime chance to rethink from first principles how the world works, and the most articulate thing the industry can say about it is a multiplier.
A friend at a prominent AI startup told me recently that their hardest problem is no longer training better models. It’s deciding which problems are worth solving in the first place —in other words, an exercise of judgment, which we suddenly discover is in short supply. For a century, the question of what’s worth doing could hide inside the question of what’s possible. Effort was scarce, any metric that economized on it received positive reinforcement. We’ve grown accustomed to the measures that no longer serve us, metrics built for a world where doing was the hard part, and now we need something new.
Changing what we focus on does not mean we lose out on productivity. Instead, it lets us use those gains to actually improve our quality of life, rather than spinning the wheels harder with no purpose. Tools like Cursor and other coding assistants can help developers spend more time solving creative problems and less time on routine tasks. But to benefit from this, we need to make a conscious choice to use the extra productivity for something more meaningful than just producing more lines of code.
An individual’s choice in the Malthusian Trap
My more self-aware friends fight the productivity treadmill on purpose. They make rules for themselves about when to stop optimizing. You can’t actually climb out of a Malthusian Trap, that’s the whole definition of one. But you can carve out a little room inside it.
A few things that seem to work.
Set your own limits. Some of the most interesting creative work today comes from people who choose to use AI less, or only in a controlled way. Back in the 70s, Brian Eno and Peter Schmidt shared this idea with their Oblique Strategies deck, which had prompts like “honor thy error as a hidden intention” to get unstuck. An artist I know runs his first concepts through a model and then does all the actual production by hand; he’s exposing himself to the Richter squeegee accident, in pursuit of a beautiful mistake that finishes the piece for you.
Open source ideas. Software is getting cheap to fork, and the moat was never really the code. Putting your work or what you know out in the open usually does more for the people around you, and for you too, than burying it in a private repo nobody can touch.
The boldest move is to rewrite the rules. I had dinner with a few founders last month. One of them barely touched his wine, and at some point, just said that when everyone can do everything, doing anything starts to feel meaningless. Nobody jumped in. Then someone else said maybe it was never about the output, it was about the people across the table from you. As cliché as that sounds and how it may have been true before any of this, we no longer have any excuse not to notice.
I watched a performance at a small theater in Brooklyn a couple weeks ago. The artist’s voice cracked during the climax of her song. In that perfect moment of imperfection, the audience leaned in. I felt my chest tighten, moved by a gentle sadness exuding from the tremor in her voice. The audience whispered afterwards, commenting similarly to the striking rawness of it all.
In a world where AI can generate flawless execution, our humanity shines brightest through our limitations, not despite them.
The contours of a 10x society are already visible in pockets of cyberspace. In the Midjourney's Discord, we see glimpses of this future: artists fluidly moving between roles, their enhanced capabilities allowing them to operate as both creators and curators, teachers and students, often simultaneously. The traditional boundaries of profession and practice blur not because they're being forcibly eroded, but because they're becoming irrelevant to how value is actually created.
When technical execution becomes trivially easy, when anyone can spin up a startup, design a fashion line, or produce a film, the scarce resource becomes knowing what's worth doing in the first place. And what’s worth doing is typically deeply subjective.
Many have asked me recently, somewhat anxiously, about how I see the future of work unfold. I start asking them: close your eyes and think what amplifies your soul?
The answers I’ve gotten were about overcoming some really hard thing like finishing a marathon, experiencing a mind-blowing hospitality experience provided at a hole-in-the-wall bar, or an imperfect but emotionally-filled performance on Broadway. I thought of this Joan Didion quote on self respect:
“To do without self-respect, on the other hand, is to be an unwilling audience of one to an interminable documentary that details one's failings, both real and imagined, with fresh footage spliced in for every screening. (...) To live without self-respect is to lie awake some night, beyond the reach of warm milk, phenobarbital, and the sleeping hand on the coverlet, counting up the sins of commission and omission, the trusts betrayed, the promises subtly broken, the gifts irrevocably wasted through sloth or cowardice or carelessness. However long we postpone
it, we eventually lie down alone in that notoriously uncomfortable bed, the one we make ourselves. Whether or not we sleep in it depends, of course, on whether or not we respect ourselves."
Art collectors instinctively know this. The most successful shows aren't the ones displaying technical virtuosity but those that wrestle with distinctly human questions.
This may be the good news for those that didn’t dare to fully lean into what they love and want to do. What if the most game-optimal play in the new system is actually to become relentlessly, unapologetically you?
It’s not just “you can just do things,” but that “you can just be you, doing things.”