I’ve revised this piece a bit after hundreds of comments and some genuinely insightful exchanges. Thank you to all who engaged in good faith. I think this piece is stronger now.
You're probably familiar with the dead internet theory. Most of what you run into online now is made by bots for other bots, and last year over half of new content on the internet was AI-generated. We’re on the infinite scroll treadmill and think we’re running outside.
I’m sure you know the feeling. You go to your normal digital haunts seeking the livewire connection of mind jousting with mind, and, where once, briefly, message boards and blogs sparked, you now find a relentless stream of slop. Silicon Valley promised an age of superconnectivity and all we got was this t-shirt delivered by a neo-serf asking for a five-star rating and a slightly larger tip.
That’s bad enough. I want to talk about something worse. The dead economy theory.
The Math
The combined investment in large-scale AI infrastructure at OpenAI, Anthropic, Google DeepMind, Meta AI, and Microsoft now runs into the hundreds of billions of dollars, with projections into the trillions over the next decade. OpenAI alone has been valued at north of $800 billion. Anthropic has yet to produce a single year of profit and carries a valuation in the same stratosphere. Numbers like that need an addressable market big enough to justify them, and the only market that big is the global labor market.
While the rest of us work out what to do with claude.md files in Cowork, the industry is pitching something else entirely. Every investor deck that promises an AI agent "doing the work of ten analysts" is describing labor replacement.
The breathless announcements about the latest models and their scorecards, the new product iterations—Tag and Design and Cowork and Dispatch—are marketing. It’s easy to lose sight of the fact that the financial model underneath needs to eliminate human cost centers at civilizational scale. Absent that, these are the most overvalued assets in the history of capitalism. The people writing the checks are not in the habit of lighting trillions of dollars on fire for a better autocomplete and longer and longer memos that nobody reads.1
One of the new rituals in a new release is the announcement of benchmark scores. We open the press releases and marvel at how intelligent these LLMs are. This one disproved the Jacobian conjecture! Behind the sheen, the companies are telling us what their game actually is. Their benchmarks are designed to test how much human labor can be displaced. OpenAI’s GDPVal benchmark measures models across forty-four occupations, from real estate broker to news analyst. Mercor’s AI Productivity Index scores them against four professional roles: investment banking associate, management consultant, Big Law associate, primary care physician. Nobody assembles a list like that to find out whether the models are helpful. An OpenAI evaluation lead told the New York Times2 that models now achieve “over an 80 percent win rate compared to human professionals” on tasks that, months earlier, no model could match. A former banker on the research team “keeps being shocked by how much of her old work the models can do.”
Let’s take them at their word and assume the technology works as advertised and AI systems can do most cognitive labor at a fraction of what humans cost. Here’s what happens:
Turn one: a company sees AI as a GLP-1 for business and replaces a chunk of its workforce with agents. Costs drop and margins expand. Oh boy are shareholders happy! Investors watch a company shed jobs and bid up the price.
Turn two: the fired workers stop making money. They tighten their belts, think twice about that elite college Junior got into. When they stop spending, the businesses they used to support lose revenue. Some of those businesses fire people too. Demand shrinks a little further with each round.
Turn three: to its shareholders’ horror, the first company discovers that its customers were, in aggregate, other companies' workers. Growth stalls. Who’s left to buy widgets or software? Efficiency eviscerates TAM.3
A prisoner’s dilemma: cutting first is always the correct move because you keep the savings, and the wreckage is, for a while anyway, distributed across the market.
Henry Ford understood, perhaps apocryphally, that his workers needed to earn enough to buy his cars. The AI economy perverts this because the entire value proposition is the elimination of the human cost center. It’s akin to eliminating the workers and expecting the cars to keep selling.
The optimists will tell you this is just productivity gains and that the economy has absorbed automation before. Agricultural employment fell from ninety percent of the American workforce to two percent and civilization carried on. David Autor’s work at MIT puts roughly 60 percent of today's jobs in a category that didn't exist in 1940. New technology makes new kinds of work.
All of that is true, but it glosses over the fact that the agricultural transition took a hundred and forty years. Carl Benedikt Frey wrote a book on the Industrial Revolution and found that wages and employment took seventy years to recover for the workers it displaced. Seventy years of stagnant pay, collapsing labor share, surging profits, and Chartists in the streets. Broken eggs and omelets, right? Frey points out that "[m]ost economists will acknowledge that technological progress can cause some adjustment problems in the short run. What is rarely noted is that the short run can be a lifetime."
It’s easy to say that the economy will normalize after a period of disturbance, especially when you don’t include that the period of disturbance can swallow generations. It can be true and horrible at once.
More recently, we lived through manufacturing going overseas. The move of American manufacturing to China took decades. We are still living through the shock of it, with the hollowing out of the manufacturing-built middle class wreaking havoc on our local and national politics. Bharat Ramamurti, a former deputy director of the National Economic Council, thinks the AI displacement could run in two years. “These companies have spent so much money developing models that there’s going to be immense pressure on them to generate revenue through quick adoption,” he told the Times.
Imagine the chaos of mass job displacement happening over a matter of months or a few years. We’d be looking at something more like 1929 than 2008. Perhaps worse. Previous automation replaced specific tasks within jobs. The power loom replaced hand weaving, the spreadsheet replaced manual calculation, etc. In each case, the technology was narrow. General-purpose AI goes at cognitive work across every industry at once.
Leontief saw this coming in 1983 when he compared human labor to horses. The US horse population grew from nine million in 1840 to twenty-one million by 1900, seemingly immune to technological change. Within sixty years of the internal combustion engine, the population collapsed by eighty-eight percent. Nobody had anything against the horses. They just became uneconomical to keep. Leontief’s point was that there is no economic law preventing the same thing from happening to humans. Elon gets it. More babies who look like him. Bostrom's existential risks include "dysgenic pressures," and you can piece together what happens to the other babies.
Acemoglu finds that only 4.6 percent of tasks in the economy are currently cost-effective to automate with AI, and he puts AI’s total productivity impact over the next decade at 0.66 percent. Goldman Sachs projected seven percent back in 2023, before we began to see how this would unfold. McKinsey projects between 0.5 and 3.5 percent annually. Someone is catastrophically wrong, and the people spending the money are not the ones with the Nobel Prize. Over ninety percent of firms surveyed in 2025 reported no measurable impact on employment or productivity despite a quarter-trillion dollars of investment. Torsten Slok, Apollo’s chief economist, says “AI is everywhere except in the incoming macroeconomic data.”
The scariest version of all of this isn’t Skynet or some extinction event. I just don’t think those are likely. Here’s the shitty scenario: AI will encourage automation and just enough job loss to be very bad for the economy, and unlike the displacement effect of previous disruptive technologies, AI won’t offer the same productivity gains as, say, industrialization. This is already happening. Between 1987 and 2017, Acemoglu found, “the displacement effect of new technologies far outweighed their productivity and reinstatement effects.” New tasks arrived too slowly to absorb the people the old tasks shed. AI looks to be worse. Firms are doing what he calls "excessive automation," killing jobs without meaningfully lowering production costs and pushing the consequences onto everyone else. Much of what's being deployed isn’t good enough to justify the displacement it causes. But you gotta hit those quarterly earnings reports with a good story.
A Political Vacuum
We’ve become lazy. We take democracy as a natural order, rather than a fragile bargain. Citizens have things the governors need: labor, tax revenue, military service, consumer spending. That’s the contract: citizens consent to accountable elected government and abide by the rules promulgated by those they elect. Power stays distributed because the people at the top need something from the people at the bottom. We cede the unilateral right to use violence to the government under these rules, and rescind that grant when the agreement breaks.
Generally speaking, empires do not fall through defeat. They go broke. Take labor out of the equation and the fiscal machinery starves all at once. The tax base erodes, collective bargaining goes vestigial because employers who don't need employees don't bargain with them, and consumer spending contracts along with the wages that fund it. Piketty’s r > g accelerates, because AI cuts the last thread connecting capital accumulation to any need for human labor. Philip Trammell has no illusions about what that means: without redistribution, "approximately everything will eventually belong to those who are wealthiest when the transition occurs."
There’s a great irony in the fact that these AI companies, hoovering up all of our collective works of art, our ideas and dreams, the tiniest mundane comment on Reddit, are also built on a public research foundation.4 I’m not arguing that Google, OpenAI, Anthropic, etc. haven’t done tremendous work. They have. The question is who was paying during the stretch when the work might have amounted to nothing. AI winters were real, and what carried the field through them was grant money. Private capital arrived after the science was de-risked and captured the return. Mariana Mazzucato’s warning is that "AI risks becoming another engine of rent extraction rather than value creation." We paid for the revolution and are being asked to accept displacement as the price of progress that someone else banks.
You can still vote, and please do, for people who understand all this and are willing to try to stop it. But we need to be honest that, unless we institute some change, what you'll be voting over is the disposition of a shrinking pool of resources while the real economy runs in a parallel system you have no input into.
The people building these systems understand this perfectly. Dario Amodei has said on the record that “[t]he balance of power of democracy is premised on the average person having leverage through creating economic value. If that’s not present, I think things become kind of scary.” Dario is just saying the damned thing: his own product undermines the material basis of democratic government. And he goes on building, headlong.
Peter Thiel, a vampiric figure if there ever was one, wrote in 2009 that he no longer believed freedom and democracy were compatible. Democratic systems produce regulation, redistribution, and accountability, all of which slow down exceptional people, like him, natürlich, who want to reshape the world. If you think you're building the most important technology in human history, oversight is an affront.
When he says freedom and democracy are incompatible, he means his freedom. Our freedom is immaterial, NPCs that we are.
Thiel’s view has spread, and it explains behavior that otherwise looks strange: the political spending, the media acquisitions, Sam Altman touring the Gulf to cut compute deals with autocratic governments. Those are rational moves once you've decided democratic governance is a legacy system to route around when it interferes.
Autocracies are structurally better customers for this technology. A democratic government that uses AI to cut its own workforce faces voters at the next election. An authoritarian one doesn't, and it gets a surveillance dividend on top of the payroll savings. The Gulf states have the capital, the centralized decision-making, and no electorate to answer to. The economics point toward whoever has the fewest ways for the governed to object.
Which is part of what's drawn the Valley to Trump. He and the people around him can be bought, and they have no particular loyalty to democratic constraint.
Restlessness
The default solution to the displacement problem reads like someone prompted GPT and asked it to ELI5: just give everyone just enough free cash that they’ll cow themselves. Unimaginable wealth for Sam and Elon, and some handwave at the leisure economy for the rest of us. The assumption is that people handed a check will find meaning in hobbies and community. They’ll paint. They’ll garden. They’ll finally write that novel.
This is ahistorical bullshit.
We don’t have to speculate about what happens when economic function disappears from communities. Just look at Appalachia or the deindustrialized Midwest. Anne Case and Angus Deaton track what they call deaths of despair: suicide, drug overdose, and alcoholic liver disease, concentrated in less-educated populations that used to depend on manufacturing or other declining industries, such as mining. Poverty alone doesn't explain the pattern. What goes with the job is social standing and the sense that the next decade holds promise for you and your children. The industries left, and the vacuum was filled with opioids, domestic violence, and a life expectancy that fell year over year in the richest country on earth.
Molly Kinder at Brookings drew the connection explicitly in Sun’s NYT piece: “Our economy grew extraordinarily and prices went down, but there were clear losers.” Globalization did this, and now the AI companies are spinning the same story. "I've interviewed so many college students who are super fearful about what the future means," Kinder told the Times, "and their narrative is exactly the same as those blue-collar guys in the heartland." The software engineer in San Francisco and the laid-off machinist in Ohio are asking the same question about what happens when the market decides their skills are worth nothing.
Permanent economic insecurity eats at people whether or not the rent gets paid, and four decades of neoliberal policy plus digital acceleration built that class well before AI showed up. What AI does is widen the door to let in the college-educated professionals who assumed they were on the other side of it. We are left with Guy Standing’s precariat, or as the guys building AI now put it, a Permanent Underclass.
Anthropic’s own research shows the problem is worse than just displacement. The use of AI actively deskills knowledge workers. Junior engineers leaning on AI coding agents didn't finish tasks faster, and they understood their own work less when quizzed on it afterward. So using Claude Code didn’t provide an efficiency gain, but it did rob the engineer of learning from the build. Perfect! The tools are eroding the expertise of the next generation while competing with that generation for jobs.
The labor market capture needed to justify their valuations would drive instability that would make the current populist moment appear quaint. Tens of millions of people in their productive years, with no economic function and no route to one, watching the richest human beings who have ever lived explain that this was necessary. Stiglitz expects AI to land on "routine white collar jobs," the college-educated desk work that felt safe while manufacturing went under. Accountants, analysts, junior lawyers, radiologists, software developers. That group has been the ballast of political stability in developed democracies. The ugly secret is that it’s when this class of workers gets fucked that revolutions happen.
Nobody wants violence.5 Still, the conditions that produce it are being engineered with extraordinary efficiency by people who have apparently never opened a history book. It’s already started. In April someone tried to firebomb Sam Altman’s home. Another attacker went after an Indianapolis city councilman who had approved a local data center. Alex Karp of Palantir told a panel that "the biggest challenge to A.I. in this country is political unrest. If I were sitting here in private with my peers, I'd be telling them the country could blow up politically and none of us are going to make any money when the country blows up."
The Boys who would be Philosopher Kings
The people building AI are true believers, in the same way that cultists are true believers, the sole people who have access to revelation. A sense of righteousness runs through Silicon Valley, from the Thiel Fellowship to the rationalist blogs to the effective altruism movement.
These people believe they are operating at the frontier of human thought, but they’d have trouble in a second-year philosophy survey. They have the confidence, but no awareness of any counterarguments.
None of which is a claim about their raw intelligence. They’re very intelligent in a narrow capacity, and like many brilliant people, mistake their narrow genius for a global one.
Let’s start with Nietzsche. The Valley loves Nietzsche, or rather a version of Nietzsche that would have made the man lose his shit and go horse-hugging much faster than the syphilis. The Übermensch gets trotted out as justification for the exceptional founder, the visionary who transcends conventional morality because he’s operating on a higher plane. Nietzsche was diagnosing the crisis of meaning after the collapse of metaphysical certainty, not writing a management philosophy for people who got rich selling advertising technology. The Übermensch is about the individual’s relationship to the creation of meaning in a godless universe. It has nothing to do with whether Peter Thiel should be exempt from democratic accountability. Nietzsche would have classified these people as the last men, the ones who mistake comfort and optimization for human flourishing. He would have fucking loathed them.
Effective altruism is utilitarianism reinvented by people who have apparently never encountered Bernard Williams, or Derek Parfit’s own agonized wrestling with the implications of consequentialist reasoning, or the two centuries of philosophical literature explaining why naive expected-value calculations produce monstrous outcomes when applied without limiting principles. The EA movement walked itself into the Sam Bankman-Fried catastrophe because it never questioned the moral framework. It assumed that human endeavors can be reduced to equations, lives into instrumentalities.
Longtermism is Parfit without the rigor. The argument has no limiting principle. It simply holds that we should optimize for the welfare of trillions of hypothetical future beings and that present-day costs are acceptable in service of that goal. Here’s the thing: you can read this to justify whatever you were already doing by positing some downstream effect that’s cosmically important.
There’s a version that cautions against an expansive reading. Bostrom’s astronomical-waste logic points at minimizing existential risk, which usually means the brake, and MacAskill has said on record that he recommends slowing the intelligence explosion. The e/acc group are purported to be the enemy camp. I see the same DNA, expressed differently. They're the gnostics to the longtermists’ early Christians. Same stew, different tastes. Both start from the same predicates: moral weight lives in an astronomical future, the decision procedure is expected value over that future, intelligence is the pivot of history, and a small vanguard is positioned to steer it. They split on the sign of the AI term, salvation against extinction, and so on whether to stamp the accelerator or the brake. This is a sibling war.6
Parfit spent a career tormented by the repugnant conclusion and the non-identity problem, and the popular longtermist literature treats those as speed bumps rather than reason for heart-stopping vertigo. The longtermerists happily took what they liked from Parfit, but MacAskill and his ilk inherited none of the discomfort.
The rationalist community rediscovers Bayesian epistemology and treats it like something new under the sun, apparently unaware that the philosophy of science has been working through these questions since the 1920s. Blog posts get treated as foundational texts. People who have never read Kuhn or Lakatos or Feyerabend think they’ve invented an epistemology from first principles and proceed to use it as the intellectual building blocks for decisions that affect billions. God, I would love to lock these guys in a room with Karl Popper and Charles Sanders Peirce.
A Philosophy 101 student who misreads Nietzsche writes a bad paper and gets a C. A billionaire who misreads Nietzsche builds a political philosophy around the misreading and funds it with the GDP of a small nation. This is fucking insane.
Logan Roy’s devastating judgment about his children is apt here: “You are not serious people.” They care about accumulation and about winning. They are not serious about the questions that matter for what they’re building: what we owe each other, what makes a life worth living, and what happens to a civilization when you remove the material basis of human agency. Those questions have occupied the best minds in human history for millennia. The Valley’s engagement with them amounts to reading Claude’s summary in an Uber ride, arriving convinced you’ve mastered the canon.
Closed Games
Intelligence and judgment are different faculties. The reason they can’t see the difference is that they mistake their narrow conception of intelligence (the kind of intelligence they have) for the totality of intelligence.
Ask one of them what intelligence is, and you’ll get facility with formal systems: math and code, the stuff on the test they’ve been winning since they were fourteen. It’s what you arrive at by looking in a mirror and writing down what you see, then announcing that’s all that matters.
It’s a kind of intelligence, sure. Notice how convenient it is, though, that the definition they landed on is the one that makes them the smartest people who have ever lived.
Look at the problems they’ve actually solved with LLMs. Go has rules, proteins have a ground truth, and a theorem either closes or it doesn’t. Every one of those domains comes with a verification function, some way to check the answer that doesn’t require asking a person, and every one of them holds still while you attack it.
Aristotle worked out the distinction they keep flattening. Techne is craft, the skill of making a thing well. Phronesis is practical wisdom, knowing what ought to be done in a particular case where no rule decides it, and it comes from living and paying attention rather than from a formula. These men have techne like nobody in history and they think techne can solve phronesis with scaling.
But the questions they’ve appointed themselves to settle have no benchmark and nobody to grade the work. They don’t seem to have any conception of this. Dario is out there saying AI will displace most white collar work, and I have to wonder whether he’s ever held one of these jobs. This work isn’t simply computation; just ask the engineers Ford cut and then had to scramble to hire back. The skill that wins a closed game—assume the answer is in there and bring overwhelming force to get it—makes a mess of an open one. When a lab cracks protein folding and concludes it now can solve political problems, judgment has a null value in the system that produced the result.
And they want to restructure civilization.
The Person in Front of You
Albert Camus broke with Jean-Paul Sartre and the French left over the most concrete political question there is: can the people alive today be treated as acceptable casualties in the pursuit of a better future?7
Sartre and the Marxists said yes. History has a direction. The revolution requires sacrifice.
Camus said no. Any system of thought that subordinates living people to a hypothetical future has already committed the foundational moral error. Once you accept that logic, there is no limiting principle. Any atrocity becomes justifiable. Any amount of present suffering can be rationalized as a necessary input to the glorious output.
This is the structure of the AI acceleration argument. The technology will eventually benefit humanity (trillions of future humans, lives of abundance and meaning we can barely imagine), so present disruption is tolerable. Displaced workers, hollowed communities, the erosion of democratic leverage, the concentration of power in a handful of private actors who have exempted themselves from the consequences of their own project: regrettable but necessary. The expected value math works out.
The founders of Mechanize, a startup whose stated mission was “to enable the full automation of the economy,” made the logic explicit: “the only real choice is whether to hasten this technological revolution ourselves, or to wait for others to initiate it in our absence.” This is determinism washing feet and doing the work of absolution. If the future is fixed, our only choice is whether to build it first. Therefore nothing we do along the way requires justification, because the destination was never in our hands. They’re making the same argument as the Marxists who sent dissidents to the gulag. It’s an eschatological movement.
Camus was ruthlessly attacked by Sartre and his toadies for insisting that the person standing in front of you is not an input to a utility function. Their suffering is not redeemed by a future state of affairs they may never see. Their dignity is not negotiable against projected outcomes. The person who exists now (who has a job they’re about to lose, a family they support, a community that depends on a functioning local economy) is the unit of account. Not humanity in the abstract. Not the trillions of future beings that the longtermists conjure to win their expected-value calculations.
Once that commitment is abandoned, the door opens to every form of rationalized cruelty that the twentieth century spent a hundred million lives trying to teach us to reject.
The entire AI acceleration project is premised on abandoning it. It asks present people to bear costs for future benefits they may never see, distributed to people who do not yet exist, administered by a self-appointed class that has insulated itself from the consequences entirely. Altman’s “universal basic compute” concedes, if you squint, that the future he's building will need a new way of distributing things. It also puts him in charge of the distributing. We’ve already tried feudalism.
Jasmine Sun reported recently that tech industry sources “expressed more extreme concern about the labor market impacts of A.I. in private conversation, but suddenly became optimists once I turned on the mic.” They know what they’re building. They know what it will do. They perform optimism in public because the alternative is admitting that the thing they’ve staked their careers and fortunes on will immiserate a significant portion of humanity. Amodei has written that Anthropic is “currently considering a range of possible pathways for our own employees,” implying that even the people building the technology may be surplus to its requirements. He framed this as compassionate. I read it as a CEO smelling recursive self-improvement just around the corner.
The Window
I don’t want to dwell on whether AI can do what these companies claim. It may well be able to, though the gap between pitch and product is currently vast. Serious economists put the productivity gains at a fraction of what the industry projects.8 Acemoglu’s finding is that AI doesn’t need to be revolutionary to be destructive. “So-so” automation, his term for technology that's mediocre at replacing workers and cheap enough to do it anyway, displaces at scale while delivering very little. The worst case is not the superintelligence everyone argues about. It's adequate software, deployed aggressively by companies chasing a stock price, cutting jobs it can't do because the quarterly incentives demand it.
Which is Ford. The requirements went into the system, the system produced worse cars, and 350 people got their jobs back nine months later at whatever it cost to get them. Ford could afford to find out. Most firms making the same bet will find out later and won't fix it.
The window for changing that is not infinite. Regulatory capture is already advanced. AI-related investment accounted for thirty-nine percent of US economic growth in the first three quarters of 2025, which gives the federal government a direct stake in sustaining the boom. Amodei acknowledges the result: "the reluctance of tech companies to criticize the U.S. government, and the government's support for extreme anti-regulatory policies on A.I." The regulator is now aligned with the regulated. Legislators cannot match the industry's expertise, and I can’t be the only one looking at the gerontocracy leading our government right now in despair. They’re not up to this.
Some interventions are known and none of them are technically hard. Public ownership stakes in AI infrastructure. Antitrust enforcement with teeth. A real tax regime on automated labor. Branko Milanovic puts it plainly: spread capital ownership more widely and tax the highest capital incomes more aggressively. What all of them require is democratic institutions willing to take on the richest companies in human history, at a moment when those companies are spending millions to beat the politicians who propose exactly this. Throw in serious talks with China about responsible AI development and we may begin to get something in place.
All of this paints a dark picture. I may be wrong, in big or small ways. But the genesis of this piece was the inkling that we’re so far deep into building this out, with AI taking such an outsized piece of our economic growth, that this can end very badly no matter how it plays out. The dead economy isn’t one where nothing happens. Plenty will happen, and GDP may well go up, because AI investment is already propping it up. The dead economy is the one where plenty happens and none of it requires you.
The inevitability is a sales pitch, and the men selling it have told you on the record that they don’t think you should have a vote on any of this. Has anyone with the power to shape this transition thought seriously about what it means for the people alive today who didn’t get one?
Fuck no.
Yes, I know that VCs are very good at lighting money on fire. It’s pretty rare, though, for the whole lot of them to join together in the kind of bonfire raging right now.
This essay relies frequently on the outstanding reporting of Jasmine Sun’s April 30, 2026 piece in the New York Times, which you can find at: https://www.nytimes.com/2026/04/30/opinion/ai-labor-work-force-silicon-valley.html
I’m not going to link it for every quotation pulled from Sun’s piece, so if a direct quotation is not cited individually, I have pulled it from Sun’s reporting.
Falk and Tsoukalas gamed this out in a paper called “The AI Layoff Trap.” It’s worth a read.
Google paid for the transformer. Eight researchers on Google’s payroll wrote “Attention Is All You Need” in 2017 and nobody should pretend otherwise. Google did not pay for the fifty years of work that paper stands on. Backpropagation came out of universities. Hinton’s group ran on Canadian public money through CIFAR across the decade when almost nobody else would fund neural networks. The attention mechanism itself was published in 2014 by Bahdanau, Cho, and Bengio at Université de Montréal on public grants, three years before Google put it inside a transformer. The EUV lithography that every advanced chip now depends on came out of a consortium built around Department of Energy national labs. The network all of it runs on was DARPA’s.
An earlier version of this essay rather blithely stated that “The public paid for the research that made it possible.” Some good faith (and not so good faith) pushback on that point led me to pull this section into a footnote and to be more scrupulously (if tediously) accurate.
The accelerationists get there by temperament rather than by argument. They’re in love with the machinery itself, and their case reads more like a recruiting poster than a philosophy. That’s a real difference, and it doesn’t rescue them. Reaching the same predicates by aesthetic instead of by syllogism means only that they skipped the step where somebody might have checked the work. I’ve reworked some of this section after good exchanges with Ebenezer in the comments.
This event, incited by Camus’s publication of The Rebel and Sartre’s Les Temps Modernes broadside attack on it, is one of the most overlooked intellectual fragmentations of the 20th century. As you might surmise, I am, and have always been, Camusian in my leanings. A good place to begin is Spritzen and van den Hoven’s translation of the vitriolic essays between Camus and the various toadies (natch) Sartre employed. I also highly recommend Aronson’s Camus and Sartre: The Story of a Friendship and the Quarrel that Ended It, Judt’s The Burden of Responsibility, and—if you can muster the French, Onfray’s L’ordre libertaire: La Vie philosophique d’Albert Camus.
Acemoglu finds that only 4.6 percent of tasks in the economy are currently cost-effective to automate with AI, and he puts AI’s total productivity impact over the next decade at 0.66 percent. Goldman Sachs projected seven percent back in 2023, before we began to see how this would unfold. McKinsey projects between 0.5 and 3.5 percent annually.
Your three turns (layoff -> spending cuts -> revenue decline) describe the demand side precisely. I'd add a fourth that's less visible: knowledge destruction.
When companies eliminate workers, they don't just lose spending power - they lose the accumulated understanding of how things work. Boeing has been rehiring retirees and contractors to recover capabilities its workforce no longer fully holds. TSMC is discovering it now - $40B factory in Arizona, but American engineers need a year of training in Taiwan because the knowledge walked out with the people a generation ago.
And there's an irony in the "80% capability" claim that keeps appearing everywhere. Even if true - where does the remaining 20% live? In the cognitive layer that only humans carry. The one being eliminated. We're trying to fix deindustrialization by pouring gasoline on it: solving the loss of human knowledge by removing more humans. Which means even the optimistic "people will retrain" argument has a problem it hasn't reckoned with: the capacity to retrain is itself being eroded by the same system.
The dead economy isn't just one where nobody can afford to buy. It's one where nobody remembers how to make. And that loss doesn't show up in any quarterly metric - which is exactly why it keeps happening.
Your Camus point lands. The person standing in front of you is not an input to a utility function - and what they know cannot be extracted from them without extracting them.
When I read what you write after I write what I wrote, I feel like we’re thinking the same thoughts but you’re reading better books. Thank you for both making me feel like I’m not crazy and educating me to think better.