“The important part isn’t the technical characteristics of the device, it’s the power relationships of the people who use the device.”
— Cory Doctorow, The Reverse Centaur’s Guide to Life After AI
Consider two freelance writers using AI.
The first, Cory Doctorow, needs to find a quote he half-remembers from a podcast. He installs an open-source speech transcription tool called Whisper on his laptop, feeds it thirty hours of podcasts, and lets it run in the background while he keeps working. An hour later he searches the transcripts, finds the quote, and publishes a better article. Useful tool. His choice. His terms.
The second, Marco Buscaglia, is commissioned by a newspaper syndicate to “write” sixty-four pages of summer guides — a job that would have taken a team of writers, researchers, editors, and fact-checkers weeks to do properly. He uses a chatbot to generate most of it. Ten of the fifteen books on his summer reading list do not exist. The internet briefly roasts him for it. He is mortified. He absorbs the blame.
Buscaglia didn’t just make a bad decision. He was set up to fail. His job was to do the work of dozens of professionals in a time frame that guaranteed errors — and then absorb responsibility for whatever the chatbot got wrong. The chatbot doesn’t get embarrassed. The syndicate doesn’t get embarrassed. Only the human does.
“The difference between Buscaglia’s AI experience and my own,” Doctorow writes, “is the difference between a centaur and a reverse centaur.”
What Is a Centaur?#
In automation theory, a centaur is a person assisted by a machine. Think of the classic image: a human head, arms, and hands — judgment, creativity, decision — atop a horse’s powerful body. The human is in charge. The machine extends their capability.
Centaurs are everywhere and always have been. A bicycle is a centaur arrangement. So is a hearing aid, a calculator, a spellchecker, a pair of reading glasses. When you use a machine because you chose to and it makes your work or life better, you are a centaur.
Doctorow has built his entire writing career on centaur arrangements: the typewriter at six, the Apple II at nine, word processors, version control, spellcheck. Each new tool extended what he could do. None of them set the pace of his work. None of them held him accountable for their own failures. He was always free to say no — to use the grammar checker or not, to go back to pen and paper when he was stuck.
That last part turns out to be the whole point.
What Is a Reverse Centaur?#
A reverse centaur is a human conscripted to act as an assistant to a machine. The body is still human — there are things the machine can’t do yet. But the mind is the algorithm’s. The human executes. The algorithm decides.
Doctorow’s reference point is a classic I Love Lucy episode: Lucy and Ethel work on a chocolate factory assembly line, wrapping bonbons as they come off the belt. As the belt speeds up, they have to move faster and faster, beyond any human capacity for careful work. They are reverse centaurs — the machine sets the pace, the humans serve it.
I Love Lucy played this for laughs. Amazon plays the horror version.
Amazon warehouse workers are timed and monitored by AI cameras that track their every movement, measure their “time off task,” and penalize them for falling short of quotas designed to exceed human capacity. Amazon warehouse workers have among the highest rates of on-the-job injuries in the U.S. warehouse sector. Many resort to urinating in bottles because the quota system leaves no time for a bathroom break. They are not the users of Amazon’s automation — they are its peripherals. They do not assist the machines. They are used up by them.
Amazon delivery drivers must sign into at least nine separate apps at the start of each shift. They are continuously scored by AI tools for their driving: lose points for braking suddenly (even to avoid a hazard), lose points for deviating from the AI’s prescribed route (even if there’s a road closure). Impossible quotas, reduced compensation if they miss them, no time to use a restroom. The van cannot get the package onto your porch. That part still needs a human. But only that part.
These workers are not being helped by technology. They are being managed by it, measured by it, punished by it, and blamed by it when it fails.
Inevitabilism: There Is No Alternative#
The tech industry has a name for the posture it strikes when confronted with the harms of its products. Doctorow calls it inevitabilism.
Mark Zuckerberg, Tim Cook, Jeff Bezos — when pressed on the damage their platforms do, they shrug. Of course you’d like privacy. Of course workers deserve bathroom breaks. But this is just how the technology works. This is what the market demands. These things cannot be otherwise.
It is, Doctorow notes, a cheap bully’s trick. Margaret Thatcher perfected it: she called her program of austerity and cruelty an iron law of economics, repeating “There Is No Alternative” so often that critics nicknamed her TINA. Her contemporary successors in Silicon Valley C-suites are TINA’s children: insisting that their abusive arrangements are out of their hands, that they are merely acting in accord with some great force of history.
Inevitabilism is a lie. The arrangement between people and their technology is always a choice. There is nothing about a warehouse that requires workers to be penalized for bathroom breaks. There is nothing about an AI radiology tool that requires the radiologist to be transformed from a professional into a liability-absorbing rubber stamp. These are choices made by people with power, dressed up as iron laws, to prevent the people without power from noticing they could demand something different.
Doctorow is a science fiction writer, and he argues that science fiction is an inherently anti-inevitabilist literature. By imagining many different possible futures — some better, some worse, all plausible — SF demonstrates that the future is not foreordained. The current arrangement is a choice. We can choose something else.
What the Bubble Is Really For#
Here is the honest math behind the AI investment boom, as Doctorow lays it out.
Analysts at firms like Morgan Stanley calculate the “market opportunity” of AI by enumerating the workers whose jobs it might displace, multiplying their wages, and applying a discount for what their bosses would pay for software that could do those jobs instead. The $16 trillion figure you hear for AI’s economic impact is essentially a running total of workers’ wages that could theoretically flow back to capital if AI could eliminate those workers’ jobs.
This means that for AI to justify the hundreds of billions being spent building it, it has to be deployed to reduce wages. That is not an incidental consequence — it is the business case.
Bosses have always hated paying workers. Wage theft accounts for a hundred times more in losses than burglaries and shoplifting combined. Companies misclassify workers as contractors to avoid minimum wage laws. The biggest Silicon Valley firms illegally colluded to prevent engineers from getting raises by poaching each other. AI is the latest, shiniest tool in an ancient fight over whether profits flow to capital or to labor.
“AI doesn’t go on strike,” an AI consultant told Gizmodo. “It doesn’t ask for a pay raise.”
Scratch any story about AI displacing workers and you find a boss whose primary headache is a worker with human needs — the need for a bathroom break, the need for a fair schedule, the ability to say “that note is stupid” and push back. The AI doesn’t do any of that. That’s what makes it appealing, and it has nothing to do with how good the AI is.
The threat of AI also functions as a wage suppression tool even before deployment: workers who fear imminent replacement accept worse conditions, drop union demands, and make concessions that shift profit from labor to capital. The hype does this work regardless of whether the AI actually performs.
Accountability Sinks#
When an AI makes errors and a human is nominally “in the loop,” something specific happens to that human. They become what finance writer Dan Davies calls an accountability sink — a place where responsibility goes to die.
The hospital fires most of its radiologists and replaces them with an AI that reviews chest X-rays. A few surviving radiologists are paid to review the AI’s output — thousands of assessments a day. Their job is to click “OK” when the AI is right and catch it when it’s wrong. But a well-understood phenomenon called automation blindness makes this essentially impossible: when your job is to review something that is almost always fine, you eventually lose the ability to spot when it isn’t. The TSA proves this daily — world champions at finding water bottles, reliably missing the guns and bombs that government red-team tests try to smuggle past them.
When the AI misses a tumor, the radiologist missed it too. The AI company is not responsible. The hospital administrator who decided to fire the radiologists is not responsible. The radiologist who clicked “OK” ten thousand times that week is responsible.
The lawyer who filed AI-generated briefs full of hallucinated case citations is responsible — not the chatbot, not the firm that pressured them to use it, not the bar association that hasn’t adapted its rules.
The programmer who failed to catch that the AI-generated code called on a nonexistent library (a “slopsquatting” attack, where malicious actors register libraries with names AIs reliably hallucinate) is responsible.
This is the reverse centaur in professional contexts: the machine sets the pace and volume of work beyond human review capacity, the human absorbs the blame when the machine fails. The machine’s mistakes are laundered through the human.
The Centaur Choice#
None of this is inevitable.
There is a centaur version of AI radiology: pair radiologists with AI that offers a second opinion. Where they disagree, the radiologist checks again. This costs more, not less — you’re paying for the AI and the radiologists. But it produces better outcomes, and the professional doing the work has genuine agency.
The centaur version doesn’t interest the boss. The boss isn’t buying AI to improve outcomes. The boss is buying AI to cut the wage bill.
This is the fracture line that Doctorow’s framework reveals: on one side, people who got to choose their AI tools and use them on their own terms — who are centaurs. On the other side, people who had AI imposed on them, who serve the machine’s agenda at the machine’s pace, who absorb its failures. The technology is often identical. The difference is power.
Where the Collective Stands#
The work of the Klamath Tech Collective is, in Doctorow’s terms, work to help people be centaurs.
Our Digital Self-Defense workshops teach people to choose their tools consciously — to understand what an app does to them, not just for them. Our Privacy 101 guides exist because surveillance tools are the clearest example of technology that serves its makers, not its users. Our UpRiv|Upcycle program puts refurbished computers running open-source software into community hands — hardware and software that the user controls, that cannot be remotely updated into a surveillance tool, that doesn’t phone home.
Doctorow closes his introduction with Gibson’s maxim: “The street finds its own use for things.”
Science fiction writers make lousy prophets, he says. But they can be pretty good technology critics. And the role of the technology critic — of the community tech collective, of the digital rights organizer, of the person who teaches their neighbor how to read a privacy policy — is to make the case that the arrangement between people and their tools is always a choice.
The tech bosses want you to think it isn’t. They are wrong. The Klamath Basin has survived for thousands of years because its people made choices — about land, water, kin, knowledge, and what to pass down. The same intelligence that knows the river also knows a tool from a trap.
Read the Book#
The Reverse Centaur’s Guide to Life After AI: How to Think About Artificial Intelligence — Before It’s Too Late by Cory Doctorow. Published 2026 by Farrar, Straus and Giroux.
Also relevant from Doctorow: his concept of enshittification — the process by which platforms degrade their service to users in favor of advertisers and shareholders — which he has written about extensively at pluralistic.net .
Want to talk about this? Come to a workshop or reach us at contact@klamathtech.diy .
Stay connected: