The machines did not “win” quietly. We built them, pointed them at each other, at the planet, and at ourselves, then called the wreckage “progress.” The Luddites saw the opening moves of that game and tried to smash the pieces before it was locked in.

They were not simply technophobic peasants swinging hammers at anything with gears. The historical Luddites were skilled textile workers in early 19th century England, reacting to specific machines that let employers replace them with cheaper, less skilled labor while slashing wages and destroying craft autonomy.¹ They sent organized threats, coordinated raids, broke particular frames and shearing machines, and were met with state violence, executions, and transport.¹ Their “aversion” was not to tools as such. It was to a new regime of power in which machines were deployed primarily to discipline labor.

That context matters, because the usual story is that the Luddites “lost,” industrialization lifted all boats, and we should stop whining and retrain for whatever app or platform is currently chewing its employees. The actual record is less tidy.

The early victory of the machines

In the early industrial textile trades, mechanization did not immediately raise wages. It did the opposite for decades. A recent reconstruction of British textile data shows that while cotton textiles became one of the largest sectors of the British economy, real wages for weavers stagnated or fell for a long period, even as output and profits surged.² David Ricardo, one of the founding theorists of classical economics, famously had to revise his views on machinery when he saw that new devices could enrich capitalists while immiserating workers.² That was not a bug in the system. It was the point.

Factory work concentrated bodies in unsafe rooms, under long hours and close surveillance. Cotton mills relied heavily on child labor, with many workers starting before age ten and working 14 to 16 hours a day in degraded air and dangerous conditions.³ Higher wages than agricultural work came bundled with shorter lives, broken bodies, and less control over time.³ The “machines” were not just physical devices. They were factory systems that reorganized space, time, and authority around throughput and shareholder return, not human flourishing.⁴

If you imagine the Luddites looking at this emerging order and deciding “no,” it begins to sound less irrational and more like basic situational awareness.

From spinning frames to automated everything

Fast forward two centuries. The basic pattern has scaled.

Modern automation is not just about replacing a handloom with a power loom. It is about turning more and more tasks into programmable routines that can be executed by machines, software, or a handful of highly leveraged humans. Work that can be routinized tends to be shifted from labor to capital, and the gains increasingly pool at the top.

Contemporary economic models find that automation generally increases income and wealth inequality unless it is coupled to strong redistribution and institutional safeguards.⁵ When machines absorb tasks previously done by humans, the share of income flowing to capital rises, the labor share falls, and inequality grows.⁵ Large empirical studies of the United States suggest that “task displacement” from automation explains a substantial portion of wage gaps and the hollowing out of middle skill jobs over recent decades.⁶

In other words, the intuition behind machine breaking, stripped of its sledgehammers, is now a peer reviewed literature: if you design technology inside a system that rewards cost cutting and shareholder value above all else, the benefits flow upward and the risks flow downward.

The AI phase of the same story

Artificial intelligence is just the latest front in this long war of substitution and control.

Current estimates suggest that generative AI could significantly disrupt the tasks of more than 30 percent of workers in advanced economies, with up to 85 percent seeing at least some fraction of their work changed or displaced.⁷ The impacts are not evenly distributed. High exposure is predicted in clerical, administrative, and “cognitive” roles.⁷ International bodies project that AI will affect around 40 percent of jobs worldwide, often widening inequality both between and within countries unless policy deliberately pushes in the opposite direction.⁸

Surveys already show that many workers expect AI to increase income inequality and job insecurity.⁹ Early data is backing that up. Recent studies indicate that AI adoption is beginning to displace younger workers in sectors like customer support and software development, with significant drops in employment for people in their early twenties compared to older workers who can leverage AI as an amplifier rather than a replacement.¹⁰ Job transformation is also gendered. The International Labour Organization has warned that AI threatens a larger share of women’s jobs, especially clerical and administrative tasks, than men’s.¹¹

Meanwhile, the stock market rewards companies that automate aggressively. Big tech firms have laid off tens of thousands while loudly celebrating their AI roadmaps. Share prices rise, headcounts fall, and a thin layer of highly paid AI specialists sits on top of a growing underclass of precarious workers.¹²

From a Luddite perspective, none of this would be surprising. They understood that machines are not neutral. They are introduced by someone, for someone. The disturbing truth is that we built economic and political structures that treat machine expansion as an unquestionable good, even when the human outcomes are obviously bad.

Consider three trends:

1. Algorithmic management
   Workers in warehouses, call centers, delivery services, and even white collar offices are increasingly managed by software. Metrics, not human judgment, decide acceptable pace, bathroom breaks, and future employability. The “boss” is a dashboard. Worker input is mostly a source of training data. This is not machines becoming conscious. It is corporations using machines to make themselves less accountable.
2. Environmental externalities
   Training and running modern AI models requires vast amounts of electricity and water, on top of already unsustainable industrial supply chains. Climate researchers and labor advocates have already warned that large scale AI could become an “engine of inequality” that intensifies both resource extraction and social strain if left to current market logics.¹³ A hyper industrial, hyper automated civilization running on fossil infrastructure is not exactly something you want in permanent growth mode on a finite planet.
3. Lock-in of technical choice
   Once you build a system that presumes constant automation, just-in-time logistics, and datafied everything, it becomes very hard to step out of it. The infrastructure shapes what is politically “realistic.” Critics sound nostalgic or ignorant, even when they are describing basic constraints like human attention, social trust, and ecological limits.

We seem happy to redesign society around what is easy for machines and profitable for owners, then shrug and pretend nothing else was possible.

Were the Luddites “right”?

If “right” means “technological stagnation would have been better,” then no. Pre industrial Europe was not some egalitarian garden. If “right” means “they correctly sensed that industrial machinery would be used to degrade labor, centralize power, and treat people as disposable inputs,” then yes, they were depressingly accurate.

Their error, in retrospect, was tactical and political, not diagnostic. Smashing machines in isolated mills could not stop the larger coalition of state, capital, and empire that profited from mechanized production. But their critique speaks directly to our situation:

• They demanded that technology be accountable to the communities it affected.
• They refused to confuse “more output” with “better life.”
• They saw that once certain standards of production and profit were normalized, resistance would be labeled irrational, criminal, or “anti progress.”

All of that applies cleanly to the world of AI, automation, and platform capitalism.

What a non suicidal relationship with machines would look like

If you take the Luddite insight seriously but do not romanticize going back to hand spinning in the dark, you get a different question: not “machines or no machines,” but “who controls them, toward what ends, and under what constraints.”

Some practical directions follow from that.

1. Use labor law and tax policy to slow the race to automate humans out of the loop
   You can tax automation and capital equipment more like labor, or give explicit incentives for augmentation rather than pure replacement. Economists have already warned that current tax structures essentially subsidize automation in a way that magnifies inequality.⁵⁶ You do not have to pretend to be surprised by the outcome.
2. Guarantee floors before encouraging further substitution
   Universal basic income, strengthened social insurance, public housing, and healthcare are not luxuries in a highly automated economy. They are survival infrastructure. When machines become dramatically more productive, the baseline for human dignity should rise, not fall.
3. Democratize decisions about where and how new systems are deployed
   Workers, affected communities, and the public should have a say before AI systems are introduced into courts, hiring, policing, education, and healthcare. That is not “anti innovation.” It is basic risk management.
4. Shrink the domains where machine logic is allowed to define value
   Not every human practice needs to be optimized, datafied, or automated. Some domains should be intentionally resistant to machine metrics: care work, education, local governance, ritual, the arts. Refusing to quantify everything is not anti rational. It is a way of protecting things that break when forced into a spreadsheet.
5. Tie AI development to planetary limits
   Any future in which machine learning models scale without regard to ecological impact is fantasy. Energy budgets, material throughput, and climate targets should set outer bounds on compute growth, not the other way around.¹³

None of this is as emotionally satisfying as burning the looms. It is also harder to meme. But it is the sort of thing that would have made sense to people who understood that the real fight was always about power, not gadgets.

The uncomfortable conclusion

The Luddites were “wrong” in the narrow sense that they failed. The mills kept running. The machines spread. Industrial logic went global and now sits in your pocket, listening to you complain about it.

They were “right” in the larger sense that counts. They sensed the arrival of a system in which:

• production overwhelms meaning
• efficiency justifies disposability
• the machines we build to serve us become the justification for treating people as if they were obsolete parts

If we go on like this, the machines will not need consciousness to dominate. They only need us to keep mistaking their expansion for our progress.

You can call that Luddite if you want. History may call it the last time workers looked directly at what was happening and tried, however crudely, to pull the emergency brake.

---

Notes

1. “Luddite,” Encyclopaedia Britannica, updated September 29, 2025; “Luddite,” Wikipedia, last modified November 2025; “Why did the Luddites protest?,” The National Archives (UK), education resource on Luddite machine breaking and motives.¹⁴
2. Daron Acemoglu and Simon Johnson, “Machinery and Labor in the Early Industrial Revolution,” working paper, April 2024, esp. discussion of cotton weavers and Ricardo’s revised views on machinery.⁵
3. “Textile Manufacturing,” History Guild, on child labor and working hours in British cotton mills; “The Rise of the Machines: Pros and Cons of the Industrial Revolution,” Encyclopaedia Britannica, September 29, 2025.⁹¹⁷
4. E. P. Thompson, The Making of the English Working Class (New York: Vintage, 1966); synthesized in Acemoglu and Johnson, “Machinery and Labor in the Early Industrial Revolution.”⁵
5. Klaus Prettner, “Innovation, automation, and inequality: Policy challenges in the race against the machine,” Journal of Economic Dynamics and Control 116 (2020); Daron Acemoglu and Pascual Restrepo, “Tasks, Automation, and the Rise in US Wage Inequality,” NBER Working Paper, 2022.⁶¹⁰
6. “AI, automation, and the future of work: Ten things to solve for,” McKinsey Global Institute, June 2018.⁷
7. Mark Muro et al., “Generative AI, the American worker, and the future of work,” Brookings Institution, October 10, 2024.³
8. Kristalina Georgieva, “AI will transform the global economy. Let’s make sure it benefits humanity,” IMF Blog, January 14, 2024.¹¹
9. “AI’s impact on income inequality in the US,” Brookings Institution, July 3, 2024; “Three reasons why AI may widen global inequality,” Center for Global Development, October 17, 2024.²¹⁴
10. “AI Is Eliminating Jobs for Younger Workers,” Wired, September 2025, reporting on a Stanford University study using ADP payroll data.⁴⁶
11. “AI poses a bigger threat to women’s work than men’s,” Reuters report on International Labour Organization findings, May 20, 2025; “Study says more women than men in Africa will likely lose outsourcing tasks to AI,” Associated Press, April 2025.⁴¹⁴³
12. “Big Tech has cut over 100,000 jobs this year – and the AI revolution is just getting started,” New York Post, August 3, 2025.⁴⁸
13. “‘Engine of inequality’: delegates discuss AI’s global impact at Paris summit,” The Guardian, February 10, 2025, on AI’s environmental footprint and inequality risks.⁴²