The End of Labor

How many times have we heard this before.

In the spring of 1589, an English clergyman named William Lee watched his wife knit. She was knitting, as far as he could tell, the way every woman in every English village had knitted for centuries: slowly, by hand, one stitch at a time. Lee had a quiet, practical mind and a slow-burning irritation. He decided there had to be a faster way. Over the next several years, in a small workshop in Calverton, Nottinghamshire, he built the stocking frame — a wood-and-iron contraption with a row of bearded needles that could produce in minutes what a skilled knitter produced in hours.

Lee was so confident in his invention that he traveled to London to demonstrate it to Queen Elizabeth I and ask for a patent. The Queen, by most accounts, refused. Her reasoning was not technical. It was political. "Thou aimest high, Master Lee," she is supposed to have said. "Consider thou what the invention could do to my poor subjects. It would assuredly bring them to ruin by depriving them of employment, thus making them beggars."

Whether the line is apocryphal or not — historians argue about it — the sentiment is real. The Queen denied the patent. Lee left England, took his machine to France, and died in poverty. The stocking frame eventually came back. The hand-knitters, of course, did not survive its return. But here is the strange part: England did not collapse into a nation of beggars. The textile industry exploded. Within two centuries, the descendants of those displaced knitters were working in mills that produced more cloth, in a year, than every hand-knitter in human history had produced in a millennium.

This is the oldest story in the modern economy, and we have been telling it, in slightly different costumes, for more than four hundred years. It is the story of the End of Labor.

The current version of the story stars artificial intelligence. Knowledge workers, we are told, are about to experience what factory workers experienced in the 1970s. The lawyer, the analyst, the copywriter, the radiologist, the junior software engineer — they are the new hand-knitters, and the model that just shipped is the new stocking frame. Goldman Sachs runs the numbers; McKinsey runs the numbers; the Wall Street Journal runs a profile of someone who lost their job to a chatbot. Every quarter brings a new study, each more alarming than the last, each phrased as though we have never seen anything like this before.

We have seen something like this before. We have seen it many times.

Ray Dalio, the hedge fund manager who has spent the better part of his career trying to flatten history into something legible, talks about cycles. Long-term debt cycles. Short-term debt cycles. Productivity cycles. The point of his framework — whatever you make of the specifics — is that the economy is not a line. It is a wave. And waves, by their nature, repeat. The peak you are standing on looks, from up close, like the highest point in the world. From a distance, it is the third or the fourth or the seventeenth of its kind.

The End of Labor is one of those waves. It rises with the steam engine and the spinning jenny, and it crests with the Luddites — those skilled textile workers in Nottinghamshire (the same county where William Lee built his machine, two centuries earlier) who smashed mechanized looms in 1811 because they could see, with perfect clarity, that the machines were going to take their wages and their craft. They were not, as the modern usage of the word suggests, irrational technophobes. They were correct. Their wages did fall. Their craft did vanish. They lost.

And the wave kept moving. Canals, then railroads, hollowed out the teamsters and the local producers who had built their lives around shorter distances and smaller markets. The reaper and the tractor emptied the countryside; in 1870, roughly half of American workers were on farms, and by 1970, fewer than four percent were. That is not a transition. That is an evacuation. The Ford assembly line did not eliminate the autoworker — it created him, but it created him as something different from the skilled craftsman he had been, a man who could build a carriage from raw lumber. The new worker tightened one bolt, all day, every day. The fear in the 1920s was not that there would be no jobs. The fear was that the jobs would be soul-crushing. That fear, too, turned out to be correct.

Then came the office machines. The typewriter, the calculator, the punch-card system, the mainframe. Clerks panicked. Bookkeepers panicked. In 1961, Time magazine ran a cover story warning that automation would create a permanent class of the unemployed. In 1964, a group of intellectuals — including Linus Pauling and Gunnar Myrdal — sent President Johnson a letter arguing that "cybernation" was about to break the link between jobs and income forever. They proposed a guaranteed minimum income. They were not crackpots. They were Nobel laureates. And they were, on the timeline they predicted, wrong.

Industrial robots arrived in Detroit in the 1970s and 1980s. Globalization and software automation arrived in the 1990s and 2000s — ERP systems, outsourced call centers, ATMs, self-checkout kiosks, the slow disappearance of the travel agent and the bank teller. Each wave produced its own literature of dread. Each wave produced real, concentrated, often devastating losses for specific people in specific places. And each wave, in the aggregate, produced more work, not less. Different work. Stranger work. But more of it.

This is the pattern. AI is not the exception. It is the latest entry in a very long list.

Which raises the question that nobody quite wants to ask: if labor was supposed to end so many times, and didn't, what was actually happening?

In 1930, John Maynard Keynes wrote an essay called "Economic Possibilities for Our Grandchildren." He was, at the time, watching the world fall apart — the Depression was tightening its grip, unemployment was climbing, the political situation in Europe was turning sour. And yet Keynes, in the middle of all this, made one of the most optimistic predictions in the history of economics. He argued that within a hundred years, the productivity gains from technology would be so enormous that his grandchildren would only need to work fifteen hours a week. The "economic problem," he wrote, would be solved. The real challenge would be figuring out what to do with all the leisure time.

We are, more or less, his grandchildren. The productivity gains have arrived. The fifteen-hour week has not.

What happened? The standard explanation is that Keynes underestimated human appetites — that as we got richer, we wanted more, and so we kept working to afford the bigger houses and the second cars and the streaming subscriptions. There is something to this. But it misses the deeper mechanism, which is not about what workers want but about what employers expect.

Here is the part Keynes did not see. When a technology makes a worker twice as productive, the worker does not get to leave at noon. The worker is expected to produce twice as much by five. The efficiency gain is not converted into time. It is converted into output. And the output becomes the new baseline — the floor, not the ceiling. A lawyer in 1985, with a secretary and a fax machine, was expected to produce a certain volume of work. A lawyer in 2005, with email and Westlaw and a laptop, was expected to produce more. A lawyer in 2025, with an AI assistant that can draft a motion in ninety seconds, will be expected to produce more still. The hours do not shrink. The expectations grow.

The same dynamic plays out in every industry. The farmer with a tractor does not work fewer hours than the farmer with a horse. He works the same hours and farms more land. The accountant with Excel does not leave the office earlier than the accountant with a ledger. She leaves at the same time and handles a larger book of clients. The journalist with a content management system does not file fewer stories than the journalist with a typewriter. She files more. Productivity, when it arrives, is rarely banked as leisure. It is reinvested as throughput.

This is why the End of Labor never ends. The labor does not disappear. It mutates. The shape of the work changes; the volume of the work expands.

Which brings us, awkwardly, to the question of why we keep believing the story anyway.

If the prediction has been wrong for four hundred years, what does it do for us to keep making it? What does it allow us to believe? What does it allow us to ignore? The things we believe most fervently are often the things that let us keep doing what we are already doing. Ideology, in this sense, is not a lie people are tricked into. It is a story people tell themselves so that the contradictions of their actual lives become tolerable. We know, on some level, that the story isn't quite true. We tell it anyway, because telling it lets us off the hook.

The End of Labor is one of these stories. Consider what it does for the people who tell it most loudly.

For the executive, it is a way of explaining a layoff as inevitability rather than choice. The work was going to be automated anyway. We are simply getting ahead of the curve. The decision to fire two hundred people becomes a kind of weather report — something that is happening to the company rather than something the company is doing. The moral weight is shifted onto the technology. The technology, conveniently, cannot answer back.

For the technologist, it is a way of giving cosmic significance to what is, most of the time, a fairly mundane optimization. We are not building a faster spreadsheet. We are liberating humanity from drudgery. The bigger the claim, the more the work feels worth doing — and the more the funding flows. The eschatology is good for the valuation.

For the worker, the story is more complicated, but no less useful. If labor is ending, then the anxiety I feel about my job is not really about my job. It is about the future. It is about a vast, impersonal force that nobody can resist. There is, in this framing, a strange kind of comfort. My situation is not the result of a particular employer's particular decisions. It is the result of history. And history cannot be argued with.

In every case, the story does the same work. It transforms a set of choices — about how to deploy capital, how to design jobs, how to distribute the gains from productivity — into a story about destiny. It takes a question that is fundamentally political and turns it into a question that is fundamentally technological. The Luddites understood this. They were not smashing the looms because they thought the looms were evil. They were smashing the looms because they understood, correctly, that the question of who benefited from the looms was a political question, and they were losing the political fight. Calling them irrational was the move that let everyone else stop arguing about the politics.

We are still making that move. Every time we say AI is "coming for jobs," as though AI were a weather system rather than a product built by specific people working for specific companies pursuing specific business models, we let the politics drift out of view. The End of Labor is, in this sense, less a prediction than a permission slip.

There is a more useful way to look at this, and it comes, oddly enough, from a statistician who spent the second half of his life giving lectures to Japanese factory managers.

W. Edwards Deming was an American who became famous in Japan before he became famous in America, which is its own kind of story. After the Second World War, he traveled to Tokyo to help rebuild Japanese industry, and what he taught the engineers at Toyota and Sony and Matsushita was something close to heresy in the West: the problem in your factory is almost never the worker. The problem is the system. If your workers are producing defects, it is because the system is producing defects, and the system was designed by management. Blaming the worker for the system's failures, Deming said, is like blaming the wave for the shape of the shore.

Apply this lens to the End of Labor and something interesting happens. The question stops being will my job exist? and becomes what is the system optimizing for, and what does that optimization do to the people inside it?

The honest answer is that modern systems — the ones the latest wave of technology is being deployed inside of — are optimizing for output. They are optimizing for output per worker, output per hour, output per dollar of capital invested. This is not a moral failing. It is just what the systems are built to do. Public companies are accountable to quarterly earnings. Private companies are accountable to investors who want returns. Even nonprofits and governments, increasingly, are evaluated on throughput and productivity metrics. The system optimizes for what the system measures.

Now drop a powerful new technology into a system like that. What happens? The technology does not reduce the demands on the humans inside the system. It increases them. The output bar moves up. The pace accelerates. The human becomes responsible for managing more, supervising more, reviewing more, deciding more. The lawyer who used to draft a contract from scratch now reviews twelve drafts the AI produced — and is held responsible for all twelve. The radiologist who used to read a hundred scans a day now reads four hundred, with the AI flagging anomalies, and is liable when she misses the one the AI missed too. The customer service agent does not vanish. He becomes the escalation point for the chatbot, handling only the angriest, most unresolvable cases, all day, with no breathing room between them.

This is the part of the story that gets buried under the apocalyptic headlines. Labor is not ending. It is being compressed. The systems are designed to extract more from each human hour, and the new technologies are the means of extraction. The human is not removed. The human is intensified.

Deming would have recognized this immediately. He spent his career arguing that you cannot fix a system by squeezing the people inside it harder. You have to redesign the system. You have to ask what it is actually for. And here is the thing about the current moment: very few people are asking that question. We are asking how fast can we deploy this? and what jobs will it replace? and how do we retrain the displaced? — all of which are downstream questions. The upstream question, the Deming question, is what is this system optimizing for, and is that what we actually want?

It is a question that has no technological answer. It is, like the question the Luddites asked, a political one.

So what do we do with the prediction? With Goldman Sachs and McKinsey and the chatbot that drafts the motion in ninety seconds?

The honest answer is that the prediction is half right, in the way that all the previous predictions were half right. Specific jobs will go. Specific people will lose specific livelihoods, and the loss will be real, and the transition will be painful, and pretending otherwise is its own kind of dishonesty. The Luddites were right about themselves. The clerks of 1961 were right about a lot of clerks. Some of the lawyers and analysts and copywriters who are anxious right now are anxious for good reason. Their particular version of the work is going to change, and some of them are not going to make the jump.

But the larger claim — that labor itself is ending, that we are about to enter some post-work future, that the great Keynesian leisure is finally arriving — is the same story we have been telling ourselves for four centuries, and it has been wrong every time. The reason it has been wrong is not that the technologies failed to live up to their promise. The technologies did everything that was promised, and more. The reason it has been wrong is that the systems that absorbed the technologies were never designed to give the gains back to the humans who worked inside them. The gains went somewhere else. They went into output, into expansion, into growth, into the next quarter's earnings. They went, in other words, into more work — just differently distributed.

This is the thing the End of Labor narrative obscures, and it obscures it on purpose, even when no one is doing the obscuring on purpose. Labor is not a fixed quantity that gets depleted by automation. Labor is whatever the system has decided it needs from the humans inside it. As long as we have systems that measure us by output, the output expectations will rise to meet whatever the technology makes possible. There is no equilibrium where the machines do all the work and we sit on the porch. There is only a moving floor, and the floor keeps moving up.

If you want to know what work will look like in twenty years, do not ask what AI can do. Ask what your boss's boss is being measured on. Ask what the board is asking the CEO. Ask what the investors are asking the board. The answer will not be how much have we reduced the burden on our people? The answer will be how much have we grown? And as long as that is the answer, the technology will be deployed in service of growth, and the humans will be deployed in service of the technology, and the work — whatever shape it has taken this time around — will not end. It will simply be redefined.

William Lee's stocking frame did not end the labor of making clothing. It changed who did it, and where, and under what conditions, and for whose benefit. The looms in Nottinghamshire did not end the labor of weaving. The reaper did not end the labor of farming. The mainframe did not end the labor of bookkeeping. The chatbot will not end the labor of thinking.

The labor never ends. It just moves. And every generation, looking at the latest machine, mistakes the movement for an ending — because an ending is, in some ways, easier to imagine than the truth, which is that the machine is not the thing acting on us. We are. We built the systems. We chose the metrics. We decided what to optimize for. The story we tell about technology is the story we tell to avoid that fact.

How many times have we heard this before. We will hear it again.