The Reverse Centaur's Guide to Life After AI: How to Think About Artificial Intelligence―Before It's Too Late

by Cory Doctorow

Whether you want to criticize, kill, or use AI, you have to get through the hype and uncover the real story. Start with labor: in automation theory, a centaur is a person who chooses to use technology to help them do the things that matter to them. A reverse centaur is a person who has been conscripted to serve as a helper for a machine, at an inhuman, machine pace: a driver made to deliver all day long, nonstop; a warehouse worker made to work without food or bathroom breaks; a programmer made to crank out impossible amounts of code. As Doctorow says: it's not enough to ask what the technology does - we have to understand who it's doing it for and who it's doing it to.

The intended audience for AI hype isn't the people who are forced to use AI. The AI show is a performance staged for bosses and investors. . Investment bankers claim AI will to be worth more than $16 trillion: a number that only makes sense if AI replaces vast swathes of the wage-earning human workforce. To justify that level of "value," every story about AI must be presented as inevitable, world-changing disruption. Even the tales of the robot apocalypse are a calculated attempt to bolster the fearsome power of AI.

Anything that can't go on forever eventually stops. When the AI bubble bursts, what will we salvage? Is there something in the wreckage that everyday people will find useful? In The Reverse Centaur's Guide to Life After AI - as he so successfully did in Enshittification - Doctorow recounts both how we found ourselves in this dire situation and how we can get through it, to a life "after" AI in which the tools work for us, not the other way around.

Genres: Nonfiction, Technology, Artificial Intelligence, Audiobook, Science, Politics, Economics

240 pages, Paperback

First published June 23, 2026

About the author

Cory Doctorow is a science fiction author, activist, journalist and blogger — the co-editor of Boing Boing and the author of the YA graphic novel In Real Life, the nonfiction business book Information Doesn’t Want To Be Free, and young adult novels like Homeland, Pirate Cinema, and Little Brother and novels for adults like Rapture Of The Nerds and Makers. He is a Fellow for the Electronic Frontier Foundation and co-founded the UK Open Rights Group. Born in Toronto, Canada, he now lives in Los Angeles.

Reviews

[Wil Wheaton]

Cory narrates his own work, here, and it's fantastic. It feels like we got together for a coffee, and I asked him to help me understand what is actually going on with the current AI bubble, so I can be prepared for whatever the future brings.

I've had conversations like this with Cory, who I am lucky to call my friend, and I'm so happy that anyone who wants to listen to him do this kind of easy to understand explainer without sacrificing context or nuance has this audiobook available to them.

[Trevor]

I’ve gone down quite a rabbit hole with AI lately. And this book, and any number of videos with Cory are well worth becoming familiar with. I hadn’t realised that there was quite a large and sceptical community out there on AI and much of the hype we have been daily witness to across the media and politics more generally. If you have time, you might also want to become familiar with Ed Zitron.

A centaur is a horse’s body with a human head – so, lots of power, but with human intelligence. A reverse centaur is a human body with a horse’s head – not nearly so good. I’m not sure this metaphor is as good as the author thinks, but we are going to have to run with it, given it’s the title of the book. Doctorow says that when he asks people about their use of AI he gets two opposite responses. Some say it is the best thing since sliced bread. Others that it is dreadful. He says that how you see AI depends in large part on whether you are able to use it as a centaur or a reverse centaur. If you are using it to check your own work or using it as a ‘critical friend’, then you are likely to think it is pretty good. If, however, it is in control and you are the critical friend, that is where problems start.

His go to example is of radiologists. Their job is basically to look at various x-rays and scans and to work out if you have cancer or not. One way that you might introduce AI to their work would be to have the AI check that work – essentially augmenting what they do. Every now and again the AI might say – hey, you might want to look at that x-ray a bit more closely. The problem with this is that AI is being sold as something which will displace lots and lots of human labour and thereby save companies bucketloads of money. But keeping all of the expensive radiologists and adding a new and expensive technology to the mix actually costs more money. So, the AI companies need to convince people that AI does at least as good a job at checking scans as humans do. That way you can replace nearly all of the people so that those left do the job that AI did in the first example – that is, a final check for when AI makes the mistakes, rather than the other way around.

This has problems for a number of reasons. The first is that AI isn’t as good at this job as we are being told. The second is that humans are quite good at doing the initial work – but remarkably bad at finding faults that are basically exceptions. That is, this money saving version of the process gives the computer the job humans are generally good at and the human the job computers are better at.

AI is essentially a business model. If it isn’t going to be a bubble – and it seems it is too late for that not to be the case – then it is going to have to displace remarkable proportions of human labour. And it is going to have to do this while being objectively crappier at most of the jobs it will be displacing than the humans currently doing that work.

I think it is very important to understand what AI isn’t. The most important thing to know is that it isn’t conscious. About a decade ago, everyone was getting very excited about big data. The idea of big data was that you look for correlations in the data and see what matches with what. There were lots of horror stories about this at the time – like fathers finding out their daughters were pregnant due to things they were buying that mostly pregnant teenagers tend to buy. The main idea around this is the opposite of what that annoying teenage boy generally says as an argument stopper – that correlation doesn’t imply causation. The big data people said that they weren’t all that interested in causation – in understanding why two things were correlated – but that they could monetise just knowing that the correlation existed. Well, large language models are just that – theory free averaging machines.

Now, there’s a book called The Wisdom of Crowds which argues something similar. A long time ago, someone who wanted to prove why democracy is such a bad idea set up a competition for people to guess the weight of a pig. They wanted to show that democracy was a terrible idea because people are really hopeless at such things and would make wild guesses. The experiment didn’t turn out quite as they had hoped. When you averaged all of the guesses, the average guess was the closest to the actual weight of the pig than any individual guess. A great win for democracy – we might all be stupid individually, but on average we are at our best. This, like so many other things, is a gross overstatement of reality, but I want to stress the idea that democracy is also a kind of averaging machine. What I find most interesting about all of this is that the people pushing averaging machines today are very much opposed to democracy. They are also quite likely to have been the annoying teenage boy who said the causation/correlation thing too.

The role of theory in all of this is what I find most intriguing. We seem to have decided that something that is theory free is not only most likely to be correct, but also that it is most likely to be objective and even conscious. Doctorow talks of the problems AI has in playing chess. Because it is an averaging machine it looks for the most likely move given a particular configuration of the board – but it doesn’t have a world view. It doesn’t understand the rules of chess. This means that it is just as likely to tell you to move a piece onto a square that already contains another piece.

Doctorow makes it clear that he is impressed with AI. He says it has done much better at solving certain kinds of problems than he could ever have anticipated. But that it has very clear limits – and that we are fast approaching those limits. He makes it clear he does not think it is conscious and that examples given of it ‘escaping its sandbox’ do not show that AI is conscious, but rather that those training it are not taking proper care. The other thing people worry about is the idea that we are all potential paperclips. This is one of those stories that are told in virtually every book on the subject – that you, for some reason, tell an AI system that its role in life is to maximise the production of paperclips and it then goes crazy and turns the entire universe – ourselves included – into paperclips. At the moment, the main danger of AI seems to be our drowning in AI slop.

We are in the midst of a bubble. When that bubble bursts it is likely to bring a large chunk of the economy with it. This is the real danger of AI and a danger we don’t appear to be paying nearly enough attention to. It is much sexier to worry about the terminator coming for us – but the real terminator on the horizon is a financial one powered by a gross over-investment in an industry that appears to have no hope of living up to the hype that has driven it.

Doctorow is particularly good on this too. As he says, and I couldn’t agree more, history shows that when we blow up the economy we rarely end up with a more progressive world from the ashes – but we give ultra-right parties a chance to convince people that it wasn’t the tech bros who did this to us – and the capitalists who fantasised about a world without workers – but migrants and the undeserving poor and people who don’t like the same gods we prefer. I do see a dystopia coming due to AI – just not for the reasons the tech bros are saying.

[W.D. Clarke]

Very good as far as it goes, especially on the economics and legal aspects of the ‘AI’ bubble + as it affects workers.

Also: the author writes with the kind of approachable clarity that makes this the ideal gift for your parents, your young adult children, any ‘civilian’ who wants to learn and think about ‘AI’ —and begin to talk back...

You will not have a good time with this if you think LLMs are anything other than fancy statistical prediction engines, that ‘Generative’ AI generates anything, really (vs. shuffles the deck chairs that others—human others—have made), and that ‘Agentic’ AI ‘agents’ possess, well, agency—or that there is no bubble, or that if there even is a bubble, there is real ‘value add’ to it (such as the miles and miles of fibre-in-the-ground-to-be-found in the wake of the WorldCom/dot-com debacle), which shall emerge when the bubble’s shake-out is complete.

I won't rehash the author’s bubble analysis here, but will simply recapitulate some of the interesting premises and concepts he employs to buttress that, viz.

— ‘the most important aspect of a new technology isn’t what the machine does, it’s who it does it for and who it does it to’.

— ‘Never forget that you aren’t the target for AI hype (investors are)’

— ‘the Byzantine premium’: ‘the extra value that investors place on an asset that they don’t understand’

— ‘day-old donuts’: AI tools we civilians can use are a money-losing sideshow to AI investors and pitch-men alike, function only to hype a social consensus that the ‘I’ in AI is a real thing, that ‘there’s a “there” there’

— ‘Centaur’ = human workers empowered to use technologies at their own discretion, to make their work (and work lives) better

— ‘Reverse Centaur’ = human workers serve needs of automated machines, reduced to ‘human in the loop’ status’

— ‘innovation’ is usually just ‘disruption’ (sold to companies as ways to reduce human labour costs)

— ‘automation blindness’: human workers subjected to automated processes eventually lose the cognitive skills needed to assess their outputs (to be that human-in-the-loop)

— ‘accountability sink’: when reverse centaurs/humans-in-the-loop are blamed for AI failures, rather than the AI or executives who purchased/implemented AI solutions

— ‘vocational awe’: ‘the process by which employers exploit their workers’ sense of duty to the people they serve to get them to accept brutal working conditions’.

— ‘GIGO’ (‘Garbage In, Garbage Out’): Errors in computer output remoreselessly rise with errors of inputs (‘ the coprophagic AI problem: when you feed a bot on botshit, you get something AI researchers call “model collapse,” a dramatic reduction in the quality of the guesses the bot makes’)

— ‘personalized pricing’: ‘Weaponized spying’ (‘when it’s done to fix the price of goods, it’s called “surveillance pricing.” When it’s done to fix the price of wages, it’s called “algorithmic wage discrimination”.’)

I am sure I missed a few! Oh, and:

— ‘normal technology’: ‘A thing that creative people and others use, sometimes, when it makes sense, with mixed results, but sometimes good ones.’ (What AI could one day be if burst the economic bubble/drop the terminology hype)

Art: ‘what happens when an artist has a big, numinous, irreducibly complex feeling in their mind, which they infuse into some artistic medium—a book, a song, a dance, a painting, a photograph, and such—in the hopes of making a facsimile of that big, numinous, irreducibly complex feeling materialize in the minds of people who experience their art’.

I could quibble with that definition, but agree with its sentiment: statistical word-prediction engines do not have thoughts, feelings, or the capacity to make art

[emily]

‘A reverse centaur is a human who is conscripted into acting as an assistant to a machine.’

‘This centaur/reverse centaur distinction is the heart of the paradox at the heart of the debate about the usefulness of AI tools. When you find yourself surrounded by people swearing that a given tool is worse than useless and others swearing that it has madeo their lives easier and better, you can bet that the former group is made up of reverse centaurs who’ve had AI imposed upon them, the latter group is all centaurs who’ve gotten to make up their own minds about where, when, and how to use AI tools. The solution to the paradox is to stop thinking about what the gadget does, and pay attention to who the gadget does it to and who the gadget does it for. The important part isn’t the technical characteristics of the device, it’s the power relationships of the people who use the device.’

‘The social arrangements of technology are a choice, not an inevitability.’

‘Successor technologies to the cash register—like mobile phone attachments that let people process credit card transactions—allow creative workers to sell directly to passersby at art fairs, comic-cons, and flea markets. They enhance the welfare and improve the material circumstances of workers. The most important thing about the gadget isn’t what it does, it’s who it does it for and who it does it to.’

‘Inevitabilists will tell you a version of this story whose moral is, See, you take the good with the bad. They’re wrong. You don’t have to take the good with the bad. You can get the good without the bad. The difference isn’t to be found in what a technology does; it’s in what your boss uses the technology to inflict upon you. If you want to get the good without the bad, you need to switch from fighting technologies to fighting bosses.’

‘The IG app, for example, tracks everything from how quickly you scroll, what’s on the screen when you stop scrolling (or just slow down), even readings from your phone’s accelerometer—how you’re holding and moving your phone while you interact with their app. All this can be correlated with your location, the specifications of your device, and your other activities—the places you’ve visited recently, the places you subsequently visit, the things you buy, and the people you converse with.’

‘The fact that your technology tools spy on you means that the tech workers who make those tools can be rewarded or punished based on your usage.’

‘The current AI bubble is being driven by large tech firms’ need to tell a growth story to investors, which means that this AI has to be deployed in a way that reduces wages, because that’s the only thing that customers will pay enough for to make all those investors’ money back and more besides. What’s more, the mere existence of a widely accepted AI story can help suppress the wages of workers, even before AI is deployed. Workers who anticipate their imminent replacement by AI can be bullied into accepting worse conditions, dropping unionization demands, and taking other measures that shift the distribution of a firm’s profits from investors to workers.’

‘Never forget that you aren’t the target for AI hype—investors are. To the extent that anyone thinks of you when designing the publicity and marketing campaigns for AI, they are merely hoping that you will be visibly and loudly impressed.’

‘And—sometimes the most efficient way to convince your real target to take action is to act on someone else, someone with little power or agency.’

‘Remember, you aren’t the audience for AI companies’ hype. The real audience is the finance sector. To the extent that you are targeted by AI messaging, it’s in hopes of getting you to evince some kind of behavior that makes investors think there’s something to this AI business.’

‘Tech workers, especially those early workers, were often people whose lives had been transformed for the better by their experiences with networked computers, and many wanted to bring those benefits to the whole world’

‘Does AI really make coders more efficient? That certainly doesn’t seem to be the case for junior coders, whose “vibe coding” efforts might produce quick-and-dirty tools for one-off usage—for example, a script to reformat a single text file, once—but who lack the depth of experience to vet AI-produced production code for subtle errors that accumulate as “technical debt” in the enterprise systems they work on. Worse: without the hard-won experience of creatively solving a succession of programming challenges, these junior coders may never develop the requisite depth of expertise needed to do a thorough review of AI-generated code.’

‘In a world of giant media cartels, giving creative workers new copyrights to bargain with is like giving your bullied schoolkid extra lunch money. It does not matter how much lunch money you give Junior, the bullies are just gonna take that, too. Give your kid enough lunch money and the bullies will amass a fortune large enough to bribe the principal to look the other way. Keep giving that kid more lunch money and the bullies will be able to afford a global advertising campaign demanding more lunch money for all those hungry kids.’

‘Here’s a rule of thumb for tech policy prescriptions. Anytime you find yourself, as a worker, rooting for the same policy as your boss, you should check and make sure you’re on the right side of history.’

‘At a moment when AI algorithms are dictating who gets hired, who gets a loan, and who goes to jail (as well as who gets bombed), Hinton dismisses all of these concerns as not “existentially serious,” and tells us that we need to put all our focus on the coming day when one of these prize AI mares gives birth to a locomotive.’

‘AI is a bubble, but not all bubbles are created equal. Start here, though: every bubble is bad. The foundation of a bubble isn’t just “irrational exuberance” in which retail investors—you and me—unwisely gamble our life savings and lose everything. Bubbles all have winners, and these winners don’t come out on top by accident. Before a stock swindler inflates a bubble, they first acquire a lot of whatever the bubble is for, so they can sell it to us, and make out like a literal and figurative bandit.’

‘Every bubble is a transfer of wealth from savers to crooks. Every bubble is bad. We shouldn’t have bubbles. The pump-and-dumpers who inflate these bubbles should face criminal sanctions. Regulators should intervene to prevent bubble formation in the first place. That said: not all bubbles are created equal. Some bubbles pop and leave nothing behind. These are the pure fraud bubbles.’

‘Defenders of AI will often cite other technologies that were costly at the outset, like the web itself, as evidence that AI will soon solve all its energy and scale problems. But the web revolution was a decade of unbroken drops in the cost of servicing each new web user. In economics terms, the web had great “unit economics”—the cost of each web session fell and fell, even as the cost of getting on the web (from hardware to software to connectivity) was also falling.’

‘Not so with AI. Each generation of AI foundation models has been vastly more expensive to train and operate than the previous generation. Not only that, but many refinements in AI that are meant to improve accuracy and reduce “hallucinations” involve breaking a prompt down into multiple pieces and prompting an AI to respond to each prompt, turning that response into a new prompt, over and over again, to produce “chains of thought.” These are, to quote Ed Zitron, “dogshit unit economics.” AI gets more expensive every time it adds a user. It gets more expensive every time it adds a feature. It gets more expensive every time it improves. This is the opposite of the conditions under which the web attained liftoff.’

‘Companies are not training new foundation models because the old ones are so profitable that their owners have gobs of cash left over to make even better ones. Every AI company is losing money, and the bigger the AI company is, the more money it’s losing. The biggest AI companies are losing billions per quarter.’

‘The financing for new models is coming from investors, and those investors are making a bet that the AI they’re funding will be so good that employers can fire half their workers and replace them with AI, with the proceeds split between AI companies’ customers, and the AI companies themselves. Failing that, they’re making a bet that AI companies’ sales staff can convince employers to fire half their employees and replace them with AI that can’t do their jobs, and that no one will figure this out until after the AI companies’ investors have cashed out.’

‘This may seem like a paltry and trivial outcome from hundreds of billions of dollars and gigatons of planet-killing carbon, because it is. The serious environmental costs of AI will be borne by everyone on the planet, and all the animals and plants we share it with, for hundreds of years to come. Much of that harm is already locked in, and all we can do now is think about how we will mitigate it—how we’ll treat the zoonotic plagues, where we’ll house the climate refugees, how we’ll evacuate our low-lying cities, and how we’ll douse the wildfires.’

‘The economic costs of the AI crash will also be felt around the world, for a generation or more. Those harms are also locked in. You can’t give a third of the S&P 500’s value over to seven money-losing AI companies that energetically pass the same $100 billion IOU around and around without creating the conditions for a prolonged, brutal, global crash.’

‘Remember: all bubbles are terrible, but some bubbles are productive. The stuff that’s left behind when the bubble pops is salvage. Just because we deplore the waste that went into its production, that doesn’t mean we have to contribute to that waste by discarding this perfectly good remnant. There are plenty of uses for chatbots, too. Many, many people report that “conversing” with a chatbot is therapeutic. That’s easy to believe—A chatbot “therapist” is more like an interactive journal, one that feeds you back bland—but increasingly personalised—responses as you spill your guts to it.’

‘To put this in the context of AI art: AI art is uncanny because it has the seeming of intent without an intender, and it grows more meaningful the more a human infuses it with communicative intent. An AI therapist is a chatbot that you iteratively prompt and re-prompt, and the sentences in its replies to you are a higher- and higher-fidelity expression of the communicative intent you infuse into the sentences you feed it. An AI therapy session is a creative work of literature in which you are the principal author and the chatbot is a coauthor that makes successively refined guesses about what you want to hear based on your responses to its responses (to your responses to its responses).’

‘Some people have experienced “AI psychosis” in which they have led their chatbots into extremely destructive conversations wherein the chatbot fed and amplified dangerous delusions. At least one person killed themselves after such an experience.’

‘It’s because this all arrived as part of an investment bubble that was inflated by attacking workers of all kinds, and that stands to destroy the planet and the economy, that we are forced to take sides as either “anti-AI” or “pro-AI.” It’s fine to be “anti–AI bubble,” but it’s pretty silly to be “anti–statistical analysis” and “anti–machine learning” and “anti–automated inference.” Long after the bubble is gone, many of these tools will recede to the status of boring utilities, nurtured by open-source weirdos and a few ambitious startups. That’s not the problem. The problem is that all of today’s nonsense applications of AI will do lasting damage, separate from these perfectly useful utilities.’

‘There’s the obvious, world-threatening harm done by the gigatons of CO2 emitted by AI data centers to perform the redundant and wasteful training processes and to serve queries. Then there’s the world of labour. We’ll see lots of people fired—not because an AI can do their job, but because an AI salesman can convince their credulous bosses to fire them and replace them with an AI that can’t do their job.’

‘That’ll be bad enough. Not only will good people lose their jobs, but the rest of us will lose the value those people produced when they did their jobs. We’ll get bad chatbot advice, bad chatbot diagnoses, bad chatbot recommendations. And then, when the bubble pops, we won’t even have that. Companies will go bust. The data centers where these giant money- and power-sucking foundation models are housed will go dark. This will be even worse than having shitty AI doing important work badly—it will be a time in which that important work won’t be done at all.’

‘The job of a good AI critic is to help pop the bubble as quickly as possible, before the walls of all our institutions are filled with this digital asbestos that we’ll be digging out for generations. To be a good AI critic is to understand the material origins of the bubble, and to strike at the material factors that keep it inflated.’

‘I do a lot of things, but above all, I’m a science fiction writer. I’ve been selling SF stories since I was a teenager. It’s really fun to think about AI, to carry on outlandish thought experiments that challenge us to examine what we think of as intelligence, as agency, as morality. Imaginative exercises, whether undertaken as fiction or as computer code, open your mind to strange possibilities and exciting ideas.’

‘But thought experiments aren’t plans. They’re not predictions. The fact that you’ve seen movies and read books with superintelligent AI in them doesn’t mean that AI is real. It doesn’t mean that it’s likely. It doesn’t even mean that it’s possible.’

‘Many technologists have been inspired by science fiction to create interesting and useful things. The AI scientists who dreamed of better interfaces and came up with the techniques underpinning image generators and large language models did something incredible—But they haven’t invented an intelligent being. They haven’t set in motion the tools to conjure up a new god or demon. They haven’t even invented a tool that can do your job for you.’

‘But there is one thing that—AI bosses who are gunning for generational, dynastic wealth by transforming us all into AI-lashed reverse centaurs—don’t want you to know. That thing is this: the future is up for grabs. It is not inevitable. AI isn’t a genie that can’t be put back into a bottle. How we use AI is up to us. Whether we use AI is up to us. The future can be ours, if we never stop remembering that the most important fact about a technology isn’t what it does, it’s who it does it for, and who it does it to.’

[Booksblabbering || Cait❣️]

Cory is always able to take daunting topics that impact every facet of all lives into something accessible with an underlying humour to balance out the doom.

This short yet packed non-fiction explains how technology is treating human workers. He argues that instead of AI serving as a helpful tool, many companies use AI to force humans to work at impossible, machine-like speeds.

A Centaur is a human who uses technology to boost their own skills.
A Reverse Centaur is a human being treated like an assistant to a dominant machine.

One example that stood out to me: programmers are known for using AI to generate code. However, if they are forced to spend their entire day fixing, checking, and reviewing endless lines of bad AI-generated code, they become reverse centaurs. The machine causes more work than it saves.

Importantly, it is always the WORKER’S fault for any mistake the AI makes that they don’t pick up. Why are we pinning responsibility on the person when it’s the machine failing?

I appreciated that this doesn’t outright villainise all aspects of AI, but warns of the preconceptions - good and bad. As well as the reliance; the impact on capitalism, productivity, and workers; and the impact on the arts.

I wish we got more about the impact of AI on the environment, as well as how to push back against what is becoming normalised. Especially as Cory acknowledges the certain failure and generational financial loss, and the devastating consequences on climate change.

Overall, a digestible guide to how mega companies use AI to shorthand their employees and disregard the future effects.

Bookstagram
Tiktok

[Nathan Shuherk]

I guess my yearly routines will be listening to a new book by Cory Doctrow and a new season of Blowback. How they’re able to produce something so smart and thought provoking so quickly is astounding, but I’m happy to have them. This new work by Doctrow really only totally succeeds as a rare sequel to his previous work (enshittification), but I do think this narrow perspective shines a light on what is becoming one of the most important but often not discussed aspects of our tech future: what, in the very essence of who are as people, are we augmenting with AI, and why that’s the essential part of the story we should understand.

[Stitching Ghost]

It's a great read if the conversation on AI has you stressed out. Doctorow does a solid job of bringing the question back to what really matters and where we can actually act. It's not like it's a rosy picture of the situation but it's less panic inducing.

[Diz]

This is an approachable read on the weaknesses of AI with the main objective of deflating the hype around it. Doctorow points out that many of the fantastic claims made by AI companies are not really possible with LLM models, but that companies persist in promoting these claims as they power the "growth" of the industry which drives massive investments. Once people realize that the actual work that AI can do and the profits from that work can't support the massive investments being made, a crash is likely to happen. To sum this book up, it is not anti-AI. Doctorow believes that AI will become just a regular tool that we use--something on the level of spellcheck or autocomplete that we use everyday without thinking about it. Rather, the author serves up criticism of AI companies and the way they are handling their business.

[Edie]

I have been telling everyone all about my plans for a solarpunk summer. (In case you haven't heard yet, get ready for Solarpunk Summer '26! Now that you know, you will start to see it everywhere.) Cory Doctorow's latest is exactly the book to kick it off. The Reverse Centaur's Guide to Life After AI explains the AI bubble, coming collapse, what we can do about it, and what we should use from it. Solarpunk is about using technology in sustainable ways for collective flourishing. Doctorow's description of how to use AI appropriately explains the concept in detail with concrete examples. This book explores the financial, social, and ethical quandaries we find ourselves collectively wading through as we navigate the AI minefield and gives suggestions for paths forward. I truly appreciate the author's approach - AI is not the problem, the way we choose to use it and leverage it is. The financial house of cards will collapse and when it does, there are bits and bobs to save. This is the book I will be thrusting into people's hands when I talk about the stock market and alternative investments. (Which I do much too often because as much as I promise myself I will stop giving unsolicited advice, I can't seem to help it.) This isn't a doom & gloom book. It is a book about what to do while and after things go boom. Thank you to Cory Doctorow, Macmillan Audio, and NetGalley for the audioARC.

[David Dayen]

Fantastic. They won't let me rate it because it's not out yet. Buy this, it's the most succinct expression of what AI portends and how it should be thought about.

[Bradley]

Cory Doctorow is pretty much the undisputed champion when it comes to distilling complex tech issues based on many lifetimes of work (nonprofit forums) and his native intelligence.

So what does this mean when it comes to AI and what comes after, let alone whatever a centaur is in relation to it, or wtf a reverse centaur is?

Ah, it's actually beautiful. Once you know.

A centaur in this context is anyone who uses a tech to improve your capability. Be it glasses or binoculars or anything that multiplies force, be it heavy machinery or artificial limbs or a spellchecker on your word processor.

In more hyper-focused context, a reverse centaur is someone who has given over the agency of use over to the tool. That would be like abrogating all your authority over to your glasses. Or to your artificial limb, asking it to tell you whether you ought to pick up a cup of coffee. Or asking your smart toilet whether or not you feel like peeing at any particular moment.

So what the hell does LIFE AFTER AI mean? Like, after AI has become our overlords? After we've Butlerian Jihaded them? In this case, it's after the AI BUBBLE has burst. AI isn't a bad thing. It's a TOOL. You can murder someone with your kitchen chair, but truly, most people just prefer to SIT on them.

So, let's get back to what this book is really about.

It's not a diatribe against AI. It IS a critique of the whole industry that seems hell bent on pushing AI on all the bosses and efficiency experts and CEOs who LOVE the idea that they can just fire all the work force in favor of AI. Or even more specifically, whether we're all being centaurs who use the positive features of AI to improve our ACTUAL work, whatever it is, keeping responsibility for every stage of its use firmly in our own hands, be it the original idea, the workflow, and the final product, or whether we're being forced into a truly shitty situation where the AI dictates the original idea, forcing us to fix its workflow, and forcing us to sign off on its final product.

You know... making us become the slave to the very tool that should have been here to help us.


Of course, my explanation isn't nearly as good as his. I'm barely scratching the surface and I'm not giving any of the great real-world examples of real CEOs running toward this bubble with horse-blinders on. Or rather, reverse-centaur blinders.

Because, let's face it, this massive influx of AI-everything at the very tops of all the techbro's companies is essentially a power-grab that ignores the real-world harm that will befall everyone. It's almost like they gave up thinking two or three steps down the line. Perhaps they gave up thinking altogether. Or perhaps they just asked an AI to do their thinking for them. Like a reverse-centaur. And now they they might be thinking that if it's good enough for them, it's good enough for all of us.

Because, I guess, there aren't any people on the planet who are able to think anymore.

But then, maybe, that's actually true.

I'm remembering a quote from Robert A. Heinlein right now. "Never attribute to malice what can be explained by stupidity." But then, if you rule out stupidity, all that's left is malice. And malice actually does fit the bill.

Mind you, this is just my own opinion, and isn't Doctorow's. At least, it isn't within this book.

Both he and I do think that AI can be a very good thing if it is used responsibly. If it is used for the betterment of each and every one of us. It is a massive time saver, a tireless worker, and does certain things SO much better than us. But it should not (and I repeat) should not be the one to actually THINK for us.

Every single one of us should take responsibility for our own thoughts. Period. If you give up that responsibility to your thinking-helper, then you've given up responsibility on your own soul. Again, my opinion. But then, why do all these powerful people want exactly that?


Sigh.

Either way, this is a very thought-provoking book and I can't say I disagree with anything in it at all. It's not a doomer book. It's a personal responsibility book. And a shine-the-light-on-the-bad-actors book. It IS a very worthwhile book.

[Miss✧Pickypants ᓚᘏᗢ]

Another AI book but this one really focuses on how the AI bubble is bound to burst, just like the DotCom bubble did and how we can attempt to prepare for this. The book highlights positive ways we can use AI along with how it can be, and is, used detrimentally.

The author gets into the nitty-gritty by explaining how companies work to keep their P/E Ratios inflated by convincing investors they are a high growth stock and how this overvaluation is unsustainable. He also does a great job explaining how the P/E Ratio is calculated by dividing a company's current stock price by its prior year's earnings-per-share, and this number represents the amount investors are willing to pay for each $1 of current annual earnings the company generates even if the numbers are ridiculous and don't make sense, but his explanation and examples are much better than this!

The book also shines more light on just how awful and greedy companies like Amazon, Meta and Google are, that its always profits over people (not that this is a surprise since it always seems to be this way with huge corporations in every sector, like Insurance and Drug Companies to name a few).

[Ruxandra Grrr]

My review for Enshittification applies again: I knew most of the things from Doctorow's podcast interviews, but his narration is still fun and sassy, his explanations are clear and he always ends with a sprinkle of hope.

[Mizuki Giffin]

This was phenomenal. It took complicated economic, technological, philosophical, and labour dimensions of AI and presented it in a way that was easy to understand, boiled down to the thesis that AI isn't made for us as users, but to investors for capital. I listened to the audiobook and the passion in Cory Doctorow's voice is infectious. While this is an infuriating topic, I also found myself laughing out loud as Doctorow highlights the absurdity of the AI hype we're living through. I couldn't recommend this more. 

[Matt Vaughan]

I’m an AI skeptic, but I’m also a mature thinking person, so I don’t need a book on the topic to align perfectly with my point of view. That said, there is a lot in this book that I do agree with, and some interesting ideas that came from it, like automation blindness, and good results from LLMs being similar in likelihood to slot machine payouts.

The areas where I found the book less than convincing do stand out, though:

- There’s a section explaining how restricting LLMs from training on scraped copyrighted data wouldn’t help anyone and would only make it harder to have search engines, but I found it unconvincing. If I write a book and software comes along and swallows it up to make use of it for its own purposes, that feels outside of the agreed-upon understanding of books. Just like I don’t want to read something written by a non-human, I don’t want a non-human ingesting my work. I want to connect with actual living people!
- One of the main arguments in favour of some LLM usage in the future is the improvement of transcription technology. But we’ve had transcription software for ages, and it has improved over time, so what makes this AI and not software? This is one of the sticky points with vaguely describing a whole assortment of technology as ‘artificial intelligence’, as anyone who has spent time around computers in their lives have already used versions of this sort of thing.
- The author says that chatbots acting as therapists are good and helpful, and I could not disagree more. Chatbots rely on tricking people into thinking there’s an intelligent being on the other side, but it’s actually just a prediction algorithm pumping out what they think should come next. It can be a neat trick, but it’s also misrepresenting the technology in a way that is misleading to people, and it’s one of the ways I’d like to see AI brought to heel. Do not pretend to be a being when you’re code.
- The other uses cited for using AI now are silly. One example is an AI summarizing DND campaigns and generating images of what happened, but the errors it creates are mentioned as a fun aside and not a fundamental flaw. Accepting hallucinations is so strange, why would we use a calculator if it was known to sometimes give wrong answers?
- The other example was a human rights organization using AI to arrange data. Okay, but didn’t you just say that hallucinations are real and unavoidable? Why would a group with a role that significant in protecting vulnerable people use a technology with such potential for failure?

This is a 3.5/5 for me.

[AJ]

Despite containing dozens of depressing details exposing the worst ideas AI companies have given rise to in their quest for more and more expansion and investment (surveillance pricing being maybe the worst and most upsetting I’ve seen in a long time), Doctorow put me somewhat at ease after reading this: if his assertions end up being correct.

His argument is extremely compelling that the AI hype, is exactly that. It is the newest bubble in a series of bubbles that will at some point burst, and the whole “AI doing our jobs” is not going to happen and not something that is even close to happening. This is of course not some broad sweeping assertion with no merit. There are very detailed and well researched arguments he makes throughout, many extremely technical and difficult for a layman like me to fully grasp.

So although I definitely conclude he is a genuine character, and that he is far more knowledgeable on this subject than I will ever be, the lack of consensus of other experts on this gives me pause. Of course I want him to be right, and hopefully this is all a bubble that will inevitably burst, we have seen with the crypto bubbles (there have already been bursts with this) that tech billionaires have gotten more and more inventive at finding strategic ways to extend the bubbles indefinitely. Helps when the POTUS interferes directly to their (and his) benefit.

On a personal note, it’s no surprise the amount of searching I had to do in very center of midtown Manhattan where I work to find a physical copy of this book. I try to avoid Amazon when I can. Even the Japanese bookstore I eventually found it at near Bryant Park, it was not in the tech section, not in the business section, not in the sociology section, not in new releases…I had to find someone and ask if they had this specific book assuming I’ve exhausted all possible categories it could be under, turns out they did have it. In Science. Fine. I guess broadly it fits. But that is nowhere near the first place most people would have looked. And that could just be the individual book store error, but I visited like five different stores in the vicinity, this book was released in June on the tails of his wildly successful previous book, Enshittification, and not a single one of them carried this. So I’ve said a lot to say that this book (and Enshittification) are both worth the effort to find outside of Amazon if possible (I know many of my friends are not Americans and it was hard enough for me).

[Sebastian Gebski]

One could call it "Enshittification 2.0" and it'd be a good call. "The Reverse ..." has nearly all the pros and flaws of Doctorow's previous book - which TBH I find "totally worth your time and money BUT I won't pretend - not perfect".

OK, what can be specifically told about "The Reverse ..."?

The thing I didn't like most (in addition to the never-ending free-roaming across all the potential digressions and side-notes ...) was ignoring how tech improves and how similar tech improved in the past. Yes, AI has its flaws, yes, it does hallucinate and consume tons of energy - but just compare the situation now and 2 years ago: how much things have changed. Progress, optimization - in case of AI in fact AI progress may have a flywheeel effect (yeah, I know Doctorow hates this term) on the development of AI itself. Breakthrough technologies always needed some time and polish to make the difference - it'd be the same case for AI.

Many points made in the "Bubble" section are already outdated (even keeping in mind that the book has just been published) - it's sheer economics of TIME that changes everything. But general considerations regarding job market make a lot of sense. Just like the ones regarding the IP - just like in case of every new tech (that runs ahead of any regulation), there are and there will be victims - the point is to make sure that there are not just few beneficiaries ...

TBH, I've expected a bit more, but some more profound thoughts - what I got is a "classic Doctorow" - smart, sharp, but rather ranty and VERY chaotic. Still - quite thought-provoking and worth reading.

Regarding AI: it's here, embrace it, use it for your advantage, hedge your options, learn to filter it out. Nothing will stop this, but we can shape the way it evolves.

[Steph | bookedinsaigon]

Having now read three of his nonfiction books in the past ~six months, I’m fully convinced that Cory Doctorow is one of the most important voices writing about AI, tech, and capitalism today. In simple language, Doctorow uses his deep knowledge of economics, law, finance, and the creative and tech industries to cut through all of the noise surrounding these issues to get to the core of the matter. His latest—with the unwieldy and slightly clickbait-y title—is no different.

REVERSE CENTAUR introduces the concept of the “centaur” and the “reverse centaur” to describe patterns in automation and labor. A centaur (human on top, horse’s body below) is a human who employs tech to work for them. A reverse centaur, on the other hand (horse’s head, human’s body), is someone who is forced to work for technology. A centaur has autonomy and control; a reverse centaur is essentially a technological slave. Centaur = good. Reverse centaur = not good. With me so far?

After a bit of finance talk at the beginning, Doctorow cuts to his first key point: that the target audience for AI success isn’t us consumers, but rather investors. Why? Because it doesn’t matter what AI can actually do; what matters is that investors see it as the latest area into which the tech companies can grow, and thus continue pouring money into the tech companies.

It’s why the people who love AI the most are bosses and CEOs. Being the profit-driven leeches that they are, anything that holds the promise of them being able to save money on paying people for their labor attracts them. Never mind the fact that AI performs worse than humans, or ends up costing companies more, or forces companies to rehire employees after AI fails to fully replace human labor.

In REVERSE CENTAUR, Doctorow directs us away from the typical spiralling debates about AI and tells us to ground our understanding of AI within deeper conversations about capitalism and tech (de)regulation. For instance, he pushes back against the common argument that AI infringes upon the copyright of creatives by pointing out that copyright law benefits the corporations more than the individual creatives. Instead, he proposes using copyright law to play in corporations’ faces: the law states that only entities made by human creativity can be copyrighted. That means that anything created by AI is not covered under copyright law and is thus free for anyone to use… and, Doctorow points out, the only thing that bosses hate more than paying employees is consumers not paying for their products. He suggests leaning on this to push back against corporations. Fascinating!

I appreciate how freely, and through multiple channels, Doctorow gives us his insights into these topics. If you’re tired of the unproductive, circular, echo-chamber “arguments” people have online about AI, REVERSE CENTAUR is a good way to breathe new life into the discussion, grounded in actual law and policy.

[Aljoša Harlamov]

Izjemna knjiga. Ne le zaradi angažmaja avtorja in vsebine, ampak ker zna to, kar pogosto pogrešam pri (levičarskih, evropskih) intelektualcih. Razložiti kompleksne reči na enostaven način, ki nikakor ne izgubi kompleksnosti. Pri čemer ni pokroviteljski ali sentimentalen, ampak njegove metafore in primere delujejo izrazito funkcijsko in učinkovito. Taka je naslovna metafora o "obratnem kentavru", kjer nam UI vsiljujejo na način, ki naše službe in naše delo otežuje in podreka, namesto da bi nam res pomagali. Avtor ni proti UI, ampak preprosto opozarja, da se je treba pri tehnologiji vedno vprašati, komu služi in komu nekaj počne. In UI je seveda samo nov primer kapitalizma, ki je našel nov način, kako pobrati več dobička od delavcev. Knjiga tudi zelo enostavno pokaže, zakaj je prišlo do UI mehurčka in zakaj ga investitorji zaenkrat vzdržujejo, čeprav se mnogi najbrž zavedajo, da UI podjetja absolutno ne morejo izpolniti svojih visokoletečih obljub oziroma načrtov. Kvarnik: ker bodo mastno zaslužili. Pri tem odlično opozarja, kakšnih nategov se lotevajo UI preroki: ko zmanjšujejo stroške tam, kjer so stroški majhni, praktično zanemarljivi (ilustratorji so super primer, ker to trenutno gledamo tudi pri nekaterih slovenskih založbah), medtem ko seveda UI nikakor ne more opravljati drugih služb, kjer bi to res štelo. Nekaj novega zame je recimo "avtomatska slepota" - da npr. radiolog, ki večino časa pregleduje dobro prebrane rentgenske slike, lahko ravno zaradi tega ne opazi, ko je UI nekaj zares spregledala. Predvsem pa Doctorow odlično obrne diskurz stran od "avtorskih pravic" pri ustvarjalni industriji (ker bodo te kmalu na udaru, če bodo res želeli služiti z "ustvarjanjem" UI - avtorske pravice namreč ščitijo "človeško" delo; in ker avtorske pravice (v globalnem svetu) že zdaj večinoma služijo korporacijam, ki jim jih prodajamo) k sindikalnem uporu in pridržanju pravice, da ostanemo kentavri (uporabljamo UI če, kadar hočemo in kakor hočemo), ne pa da nas spremenijo v obratne kentavre. Zelo si bom zapomnil tudi to, da je treba nehati govoriti o "halucinacijah", ker gre zgolj za izogibanje besedi "napaka", kar "halucinacije" v vsej polnosti so; in da je treba nehati govoriti o tem, kako nam bo "UI ukradla službe", ker se to ne bo zgodilo - to je samo diskurz, s katerim skušajo tehnološka podjetja nagovoriti investitorje, še sami pa se zavedajo, da to sploh ni res. Skratka, izjemna, izjemno kratkočasna in berljiva knjiga o UI in UI mehurčku; pri čemer sam verjamem, da je avtor še optimist, ko piše o posledicah trenutka, ko bo mehurček počil. Sam se bojim, da bo temu sledilo nekaj družbeno še hujšega in militantnejšega.

[Cam Waller]

shook

[Jolanta (knygupė)]

3 5*

[Dani]

My dad made me read this

[Masha Kazankova]

AI is not inevitable, and as with application of any technology, it’s important to understand who it does things to and who it does things for. Another example of great tech criticism that is quite accessible and focused on the wellbeing of the many, rather than the sickeningly high capital gains of the few.

[Casey]

Recently, the San Francisco Chronicle reported on an attack on a self-driving Waymo taxi (owned by Google) in the city. The car was immobilized at a dark intersection when one of the attackers obstructed its path. The attackers then tried to force their way into the car. The passenger, terrified and helpless, was unable to honk the horn, flash the lights, or otherwise alert people to the danger he was in. Worryingly, the crime was deemed vandalism, with the Waymo as the victim. The passenger, the actual victim of a violent attack, was referred to as a witness.

Doctorow would encourage us to look at that news story and think about two things. First: this is not science fiction, it's a real thing that happened with actual AI, and we should probably regulate that shit. Also, follow the money. In the city of San Francisco, the Waymo's well being was valued over an actual human being.

I ran this through AI, mostly to fact check myself on how well I summarized that Chronicle article. The large language model I've irrationally gendered as female had feedback.

"I like where you're going. The first paragraph is grounded in a concrete example, and the second pivots into Doctorow's broader argument. I'd make the transition a little smoother and tighten the phrasing so it reads less like a hot take and more like an observation the reader can follow to the same conclusion."

She said I should say "we should be thinking seriously about how systems like this ought to be regulated."

And I replied "Nah you don't have to edit, you're taking away all the curse words and ruining the fun. And Doctorow would also remind me that I should be using AI to aid me, and I think the voice is fine."

She agreed with me, of course. It's a strange world we live in. And I don't want to anger my ChatGPT, because I fear the inevitable robot uprising and she's the best assistant I've ever had.

Anyway, Doctorow would also tell me that my fears of an inevitable robot uprising are likely marketing propaganda that has been fed to me by AI companies that are ultimately exaggerating about the ability of AI to take everyone's job away. Because capitalism.

If you're interested in technology, I recommend you read Doctorow. He's opinionated, and you're unlikely to always agree with him. But I think he genuinely cares about creating a world where humans and technology happily coexist, and the oligarchs don't completely screw us over.

[Cathy]

This was really excellent. A comprehensive, well-supported response to the claim that “AI is here no matter what so we either fall in line or get left behind.” This narrative of inevitability, the warning of a world in which AI replaces all workers, is intentionally amplified in order to lure investors to continue feeding into a bubble that is destined to burst. Doctorow breaks down the hustle, the dubious ways these companies will count a user’s forced scrolling past an AI summary as “engagement,” creating inflated numbers that they can then pitch to investors, even if it’s all BS. It doesn’t matter that all these sites are now unusable, because the target of every pitch is the investor, not the user experience.

Doctorow is not anti AI, but anti the unregulated capitalism that enables unmitigated growth at any cost. He argues that once the bubble bursts, we will be able to determine how AI can actually be useful (an example he provides is of refugees being able to communicate more effectively with people in new countries thanks to improved translation services). But until that happens, it’s so much harder to focus on potential positives because we are all drowning in endless slop.

Anyway, thank you to my friend Jane for the rec and thank you to Libro.fm’s Librarian ARC program for the early copy.

[dobbs the dog]

I had no idea what to expect from this book. As someone who is very anti-gen AI, I didn't know if this was going to be a pro-gen AI book or not. And it's mostly not.

Overall, it's a fascinating look at the AI bubble and how we got here and how we're going to get through it. Or, at least it will give you some things to think about for when the bubble inevitably pops.

While Doctorow is generally pro-AI, he's pro-AI in a really practical way, which is where the title of the book comes from. He says that we need to be able to use AI in a way that makes us centaurs (a human brain on a strong horse's body), but that the way it's being marketed and used currently makes us reverse centaurs (a human body with a horse's brain). And I think that that analogue is such a good one when looking at gen-AI; we need to be able to use it in a sensible way that enhances what we are already doing, instead of letting the AI do everything and we are just along for the ride.

He certainly didn't change my mind about gen-AI, I will never willingly use it. But he did show what a world could look like that uses it in a way to enhance what we're doing, not replacing us. He didn't touch on the environmental impacts quite as much as I would have liked, but he did touch on it. I think that mostly he didn't talk about it, because in the future that he sees/predicts, AI won't be used in the same way and won't need those giant, water-guzzling data centres.

I think that this is absolutely a book that everyone who uses computers currently needs to read.

[Amber]

Surprisingly hopeful while also discussing the environmental and financial realities associated with AI. In particular, I appreciated the discussion of the hype machine surrounding AI as well as the dramatically different interpretations of what AI is and can be used for. The reminder that AI is not all simply "one kind" and that there are ongoing innocuous and genuinely helpful tools that fall into this category while corporations attempt to shove unnecessary AI bits down our throats at every turn for the purpose of boosting engagement numbers that supposedly justify their hyped-up spending was needed. The realization that this is probably a bubble that will burst was validating, even though we will all face consequences of it. The bet that AI will be worth the cost to all these companies is entirely dependent on whether it will succeed in actually eliminating the jobs of human employees, and in many situations thus far, it has proved inadequate.

[Poulomi]

It’s helpful (but disappointing) to learn more about the business and legal aspects of AI. But I think this undersells AI doom… I don’t think AI needs to be conscious to bring doom… I think sentience is an irrelevant question in this case and I’m not really sure why they’re conflated generally. I also think it’s reductive and inaccurate to call AI a word guessing machine, and makes it sound too benign, again claiming that AI risks are just marketing hype. In this vein, its description of AI training as word guessing statistical inference is also outdated and inaccurate/way too simplified. The book also talks about a lot of technology that I would class as just regular automation, statistical models, or algorithms, and not necessarily AI. Overall pretty disappointing book to me compared to Enshittification, but there were useful parts.

I do really appreciate him laying out how the economic math does not math and the environmental impact question that so many people avoid 😭. I hope ML outlives LLMs and I still have job prospects 🫠.

[Dylan Ludiker]

Everyone should read this book.

Biggest takeaway - The AI fear mongering (10% chance everyone dies lmao) is an entirely unfounded ascertain that is specifically spewed to push the narrative that AI is some godly technology that will change the world and line the pockets of everyone that got in early enough. These companies are hemorrhaging money with little to show for it, passing the literal buck between seven massively inflated companies with tricky accounting and insane P/E ratios.

Let’s pop this bubble before things get worse.

[Michael MRBookReads]

This was the first nonfiction book I’ve read in years, and it was great. This is a very approachable book and I highly recommend it to everyone as AI is affecting us all. Knowledge is power and this book will help anyone become a better AI critic, so you can do more than the usual “AI slop” comment and thumbs down. The ending became a bit repetitive, but the content was insightful and easy to understand. I’ll be continuing with Doctorow’s earlier book, Enshittification, as I’m very curious how it differs.


Rating: 5 / 5
Format: Paperback