Luddism gets a Gen Z reboot in Manhattan

The chant went up in a tight cluster on Manhattan's west side at roughly 21:32 UTC on 10 July 2026: "No Gemini, no GPT, no Grok, no Claude." A self-described Gen Z anti-tech contingent had gathered for an event billed as the "Summer of Ludd" festival, a one-day revival of the most discredited political label in British working-class history. The reference is not accidental. In 1811–12, framework-knitters in the English Midlands smashed industrial stocking-frames they blamed for starvation wages; the British state responded with a show trial, a mass execution, and a fabricated folk villain named Ned Ludd. Two centuries on, a younger cohort is reclaiming that name against a different set of machines — large language models developed by Google, OpenAI, xAI and Anthropic, named in the chant by their consumer brand.

Luddism, as a political inheritance, has always been an awkward fit for an American audience that prefers disruption to retro. But the festival makes a serious claim that deserves to be read seriously: that the latest wave of generative AI is being installed into classrooms, call centres and creative pipelines faster than its labour-market effects can be audited, and that the public conversation is being managed by the same handful of firms building the systems. That critique has older and sturdier roots than the Luddite label. The interesting question is not whether the chant was sincere. It is whether the politics behind it can survive the branding.

The framing the chant accepts

Every loud movement adopts the vocabulary of whatever it is reacting against, and the Summer of Ludd festival is no exception. The four names in the chant — Gemini, GPT, Grok, Claude — are the consumer-facing fronts of an industry that has consolidated around a small number of well-capitalised labs. Public coverage of those systems has tended to follow the labs' own product calendar: a release is covered, a benchmark is reported, a controversy about safety is aired, and the cycle moves on. Within that cadence, structural questions about training-data provenance, model-evaluation independence, and the labour of the contractors who label the data, get less column-inches than the products themselves.

The festival's organisers are borrowing from that grammar on purpose. By naming the four systems in a single chant, they collapse an entire industry into one easy target. The tactic is rhetorically efficient and politically thin. Gemini and Claude come from companies with safety teams and on-the-record concerns about model misuse. GPT, the flagship of OpenAI, has spent the last three years arguing publicly with its own board about how fast to commercialise its frontier research. Grok, xAI's product, sits inside a social-media platform whose content-moderation choices are a separate and louder controversy. Treating them as interchangeable erases real internal disagreement inside the AI sector about safety, deployment pace, and corporate governance — disagreement that, ironically, is more visible inside these companies than in much of the press coverage around them.

The structural argument underneath

Stripped of the Luddite costume, the case the festival is making has three parts, and all three predate generative AI.

The first is about consent. Generative models are trained on corpora that include copyrighted text, artwork and code, much of it scraped under licensing terms that did not anticipate the use. The legal status of that training is being tested in courtrooms from New Delhi to San Francisco, with outcomes that will reshape which industries owe which industries money. None of the labs contest that the training happened; they contest the legal characterisation. The festival's organisers point to this gap between what the industry calls "fair use" and what working creators experience as expropriation, and the argument does not require a Luddite frame to land.

The second is about labour. The human workers who label, red-team and rate model outputs — often via subcontractors in the global south, often on piece rates — are the invisible column of the AI economy. Reporting by American and British outlets over the past two years has documented platforms paying sub-USD-2 per hour for tasks that include classifying graphic content. Whether those conditions are a temporary phase or a permanent feature of the industry is the most consequential open question in AI policy, and it is the one least likely to be settled by a chant.

The third is about concentration. Four frontier labs, a handful of cloud providers, and a small number of chip designers sit on top of the stack. The economics of training a frontier model have compounded upwards, leaving the field with fewer independent entrants each funding round. That concentration shapes which languages, dialects and cultural contexts the next generation of models will be fluent in — and which they will be silently inattentive to. The festival does not press this point in its slogan, but it is the argument most likely to outlast the festival itself.

What a revival of Luddite politics actually costs

The original Luddites lost. They lost decisively, in fact: the British state executed or transported more than fifty of them, the framework-knitting industry collapsed anyway, and the machines they smashed became the backbone of an industrial economy their great-grandchildren would inherit. The historical record is not kind to machine-breaking as a tactic, and it is even less kind to the assumption that the technology being opposed can be simply refused. Cotton mills displaced handloom weavers; the displaced weavers did not return to a pre-mill economy. AI systems that demonstrate a 30 per cent productivity lift in customer-support workflows will not be voluntarily decommissioned by the firms paying for them.

What the original Luddites did get right was the diagnosis that the cost of new machinery is borne by people who do not own it and did not choose it. That diagnosis survives every technology cycle. The Summer of Ludd festival would do better to spend less time on the costume and more on the question of who bears the next round of adjustment costs — and through what institutions they will be compensated, retrained or given a vote.

Stakes and a date to watch

The near-term test is not in Manhattan. It is in courtrooms and in union halls. The first wave of copyright suits against the major AI labs is now scheduled for substantive hearings through the remainder of 2026, and the judgments will determine whether training on copyrighted material remains a de facto free input for the industry or becomes a recurring cost on its balance sheet. On the labour side, contract negotiations covering thousands of data-labellers at two of the larger subcontracting platforms are expected to conclude before the end of the year. Either of those outcomes will do more to shape the next decade of AI deployment than any number of chants at a street festival.

If the festival returns next summer — and the marketing logic of a branded recurring event suggests it probably will — the test of whether it has matured will be whether the slogan has acquired a referent. "No Gemini, no GPT, no Grok, no Claude" is a banner. It is not yet a programme. The gap between the two is the space where AI politics will actually be fought over the next eighteen months.

Desk note: Monexus framed the festival as a cultural event with political undertones, not as a bellwether of mass sentiment — coverage on Polymarket's wire noted the chant but did not quantify attendance, and the broader polling picture around AI sentiment in the United States is more ambivalent than uniformly hostile. We have leaned on the structural arguments the festival gestures at rather than on its costume.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://x.com/polymarket/status/1944985600000
• https://en.wikipedia.org/wiki/Luddite
• https://en.wikipedia.org/wiki/Framework_Knitters%27_Company
• https://en.wikipedia.org/wiki/Generative_artificial_intelligence