A Question That Deserves a Better Answer
A commenter named KevinM asked the question every AI panic eventually reaches, and he asked it in a blog comments thread, with definitions and a chain of logic. He deserves an answer better than a dismissal.
The thread ran under a post on Watts Up With That on August 5, 2026, about what AI can do for us after seventy years of development. Another commenter, gyan1, had written that AI’s main effect will be productivity and efficiency, and that GDP is going to explode when these systems scale. KevinM replied with a definition and a chain:
“Gross Domestic Product (GDP) is the total market value of all final goods and services produced within a country during a specific time period.”
“IF employment ->0 then income ->0 then demand ->0 then price ->0 then production ->0 so GDP ->0”
His closing question, in his own words: to use GDP as the metric for AI success, you must argue either that AI will increase employment, or that income will be separated from labor. His parenthetical names the mechanism he fears: “social credit score?”
Elsewhere in the same thread, a commenter named jtom made the standard reply: the industrial revolution reduced manual labor, created new jobs, shortened the working week, and raised wages. KevinM’s question deserves more than the standard reply. It deserves the mechanics.
This is a good-faith question from a sharp reader, and it deserves a good-faith answer. The conclusion is wrong, and it is wrong in a very instructive way. Every link in the chain breaks, and each break teaches something about how economies actually work.
What He Gets Right
Let us begin with the concession, because it is substantial. KevinM is right that GDP is a terrible metric for AI success. GDP counts final goods and services at their market prices. If AI makes almost everything nearly free, nominal GDP would fall even as the amount of stuff people actually enjoy explodes. The best possible outcome of AI - a world where the essentials cost almost nothing - looks like failure in nominal GDP.
That is a real problem with the metric. It is not, however, the problem KevinM identified. The chain’s conclusion is right. The chain is wrong, and the difference matters.
Employment Is an Input, Not the Output
GDP counts final goods and services - things people actually want, actually use, actually enjoy. Jobs are not the output of an economy. Jobs are the cost of producing the output. The purpose of a factory is not employment. The purpose of a factory is goods.
A world in which machines produce everything is not a world of zero GDP. It is the point of the machine - the point since the first loom replaced the first pair of hands. Every machine ever built exists to do work so that people do not have to. That is not a bug that terminates the economy. It is the very definition of progress.
The Nordhaus study we built the 2% Rule on shows why. Entrepreneurs capture about 2.2 percent of the surplus their innovations create; the other 97.8 percent arrives as lower prices, higher wages, and better products. The people who build the machines keep pennies. The rest of us receive the output of the machine in the form of cheaper everything. We have examined what happens to the jobs which are displaced by machines in Creative Destruction. The short answer: they disappear in the task, and reappear in the task’s shadow.
The Price Channel Is the Good Side
The chain treats “price ->0” as catastrophe. Falling prices are the good side. They are the delivery mechanism for the 97.8 percent.
Productivity-driven deflation is the good kind of deflation. When the cost of making something falls, competition forces the saving through to the buyer. The good that cost a week’s wages costs a day’s wages, then an hour’s. Real incomes rise even when the paycheck does not, because the paycheck buys more. This is the seen and the unseen in one move: the visible line is the falling price, the invisible line is the rising standard of living.
The commenter’s chain accidentally proves the strongest case against GDP as a measure of human welfare that exists. If AI succeeds beyond anyone’s dreams - if the price of everything useful collapses toward zero - the GDP statistic falls apart while the world gets richer in the only sense that matters. A metric that cannot see the best possible outcome is not measuring the right thing.
The Lump of Labor Was Never Real
The fixed-pie view of work has been wrong for two hundred years, and the wrongness is the entire history of rising wages.
In pre-industrial societies, the overwhelming majority of the workforce - three-quarters or more, and closer to nine in ten in the poorest places - worked the land. In the United Kingdom today, agriculture employs about one percent of workers, per World Bank data. The machines did not shrink the world’s work. Total employment in Britain is vastly higher than it was in 1800, and so is the population it supports.
Every wave of machinery destroyed jobs in the task and created jobs in the task’s shadow - jobs nobody could foresee, because nobody could see the task’s shadow in advance. The farmhand could not imagine the railway porter, the typist, the web designer. The historian of that transition, Robert Allen, named the ugly middle period. Real wages stagnated for roughly four decades while output per worker rose, in his paper “Engels’ Pause”, published in Explorations in Economic History in 2009. The episode is named for Friedrich Engels, who documented the suffering.
And then the long run arrived. The world the machines built: in 1820, roughly nine in ten humans lived in extreme poverty; today, fewer than one in ten, on Our World in Data’s long-run series. The poor have been the biggest beneficiaries of the machines, as we showed in Who Actually Helps the Poor?. Hobbes called pre-industrial life nasty, brutish, and short in Leviathan, in 1651, and he was describing the norm for most of human history. The machines ended that norm.
Income Was Separated From Labor Long Ago
KevinM’s binary assumes that income separated from labor is a science-fiction premise requiring a score. It is the oldest feature of modern economies.
Capital income - dividends, interest, rent - absorbs roughly a quarter of national income in the rich countries, on the estimates collected by Thomas Piketty and the World Inequality Database. It arrives with no wage attached. Pensions are deferred wages: a pensioner receives income without working, by design, and nobody runs a credit score to permit it. A child with a savings account earns interest without labor. The separation happened, is happening, and requires no apparatus.
And automation already bought leisure on a scale the Luddites could not imagine. The working week in Britain fell from roughly 60 hours in the mid-nineteenth century to about 40 today, and annual hours have nearly halved since the 1870s, per Our World in Data. Machines buy free time, not just cheaper goods. That, too, is income separated from labor - delivered as hours, and requiring no scorekeeper.
The Luddites Were Half Right
This is where the argument needs its history, because the history is the evidence.
In November 1811, in Nottinghamshire, stocking weavers formed a secret society. Their nominal leader was Ned Ludd - a fictitious figure, styled Captain, General, or King Ludd, depending on the handbill. Their trade was being destroyed by wide frames: knitting machines that let one man do the work of several, at lower quality and lower pay, in the middle of a war with Napoleon and the economic misery that came with it.
In targeted raids, at night, they smashed more than a thousand of the frames. The destruction spread to the textile districts of Lancashire and Yorkshire. The government’s answer was overwhelming: 12,000 troops, and a law making the breaking of frames a hanging offense. There were exemplary hangings and transportations. The outbreaks sputtered until 1816 and then stopped. The frames stayed.
To be honest, the weavers were not wrong to complain that it hurt. Real wages in Britain were roughly stagnant from about 1800 to 1840 - the Engels’ pause - while the rich got richer on the machines’ output. Four decades of flat pay is not a rounding error. It is a generation’s worth of distress, and the frame-breakers lived inside it. Anyone who tells you the Luddites were simply stupid has not read the wages series.
They were wrong about the enemy, though. Between 1840 and 1900, British real wages roughly doubled - a rise of about 123 percent, on the same historical series that documents the pause. The machine did not end the weavers’ world. It ended the world in which weaving was the only work a weaver could do. The Luddites’ descendants - the ones who did not hang - became the richest working class in history. Half right: right about the short run, wrong about the long run, and the second half of that sentence is the part that determines everything.
The Academic Blessing
The same half appears in academic dress this week. On August 5, 2026, Brian Lewis, a professor of history at McGill University, published an essay on The Conversation about who the Luddites really were. The history is careful and mostly fair. The Industrial Revolution is described as a series of “traumas.” The frame-breakers are rescued, in E.P. Thompson’s famous phrase, “from the enormous condescension of posterity.” Their violence is presented as Eric Hobsbawm’s “collective bargaining by riot.” And the essay blesses the modern heirs: neo-Luddism, Lewis writes, is being reclaimed by people concerned about working conditions, society, and democracy in an era of AI expansion.
Be generous, because generosity is owed. Lewis is right that the Luddites were not fools and not simply enemies of progress. They were skilled artisans defending a way of life, with no unions, no vote, no welfare state. Their grievance was real.
The mistake is the conclusion, not the sympathy. Lewis measures the transition and calls it the destination. The decades of the Engels’ pause become the permanent character of industrial change; the 123 percent wage rise of the following sixty years goes unmentioned. That is the same error as KevinM’s chain, in an academic gown: take the painful decade, assume it is the permanent state, and bless resistance to the machine on the strength of the assumption. The neo-Luddites of 2026 are being handed the weavers’ pain without the weavers’ payoff, as if the story ended in 1816.
The False Binary
Strip the argument down and it is a forced choice: either AI increases employment, or income must be separated from labor by a social credit score. The third path is the one history actually took. Prices fell. Hours fell. Ownership spread. Work changed shape. All four happened between 1816 and 1916, and all four are happening now.
Run the question KevinM did not run. If income must be separated from labor, who should administer the separation? The market’s version - falling prices, dividends, pensions, cheaper everything - needs no scorekeeper. It is the most democratic mechanism ever built: every purchase is a vote, cast continuously, by everyone, as we argued in The Free Market Is More Democratic Than Voting. The state’s version - a social credit score - requires a new apparatus to decide what each person deserves, which is the Information Problem in its purest form: no authority can know it. And the people who benefit from the apparatus are the people who administer it. The panic serves the same coalition that always profits from panic, the pattern we traced in the grievance machine.
What the Machines Make for Free
While the economists argue about the end of the economy, the thing itself is already at work. Sam Berlin’s “AI Is Free and Actually Useful” on The Basic Hub is the practical counterpoint: what AI does today, for free - drafting emails, explaining things you do not understand, summarizing long articles, planning trips, generating ideas. No world-ending stakes. No social credit score. An hour a day returned to the user, at a price of zero.
That is the 97.8 percent of value arriving in real time. The panic is a debate about the transition. The free hour is the destination.
The Lens
The best outcome of AI looks like the end of the economy if you measure the economy by its inputs. Measure the outputs. The Luddites were half right: the transition hurts, and the hurt is real. They were half wrong: the machines were not the enemy, and the proof is in the two centuries of wages, hours, and improved poverty statistics that followed.
Next time someone tells you the machines will take everything, ask what the machines will make - and who it will be cheaper for. The answer to the second question is the whole enchilada.