This is polemic dressed as explainer, and the disguise mostly works. Doctorow isn't interested in whether large language models are impressive parlor tricks or genuine breakthroughs; he's interested in who cashes the check when a company tells investors its product will replace a warehouse floor or a legal department. That's the real subject of this book, and once you see it, the AI debate looks different.
The argument, stripped to one sentence: AI is worth sixteen trillion dollars only if it eliminates enormous swaths of paid human labor, so the industry's incentive is to sell that outcome as inevitable whether or not the technology can actually deliver it. Doctorow builds this case less through technical critique of neural networks and more through the economics of valuation itself, tracing how Silicon Valley has run this exact play before, with crypto, with the metaverse, with dot-com stock in 1999. He knows the pattern because he's watched it up close for two decades as an activist and columnist, and the chapters that dig into how a company's stock price gets tethered to a story rather than a balance sheet are the book's strongest.
The centaur and reverse centaur framing is where the labor argument gets its teeth. A centaur uses a tool to become faster and better at a job they still control. A reverse centaur is controlled by the tool: the Amazon picker paced by an algorithm, the delivery driver routed by an app that punishes bathroom breaks, the coder measured by lines generated rather than problems solved. Doctorow's point is that most current AI deployment sorts workers into the second category, and that this isn't a bug tech companies are racing to fix. It's the business model.
There's a chapter midway through that walks through several real corporate rollouts of AI tools, and it's the most useful stretch in the book precisely because it's concrete. Doctorow isn't allergic to nuance here; he name-checks genuine productivity gains alongside the failures, and he's upfront that he uses AI tools in his own writing process. That admission matters. It keeps the book from reading like a Luddite tract and lets the critique land as economic rather than moral.
Where the book strains a bit is in its brevity. At under 250 pages, this reads more like an extended essay collection than a fully load-bearing argument, and a few sections feel like columns stitched together rather than chapters building on each other. Readers wanting deep technical grounding in how transformer models actually work, or a granular policy roadmap for regulation, won't find it here. Doctorow trades depth for velocity, and mostly that trade pays off, but the last third occasionally repeats points already made in sharper form earlier.
Why you should read
- Readers skeptical of AI hype cycles
- Fans of Doctorow's essayistic, blog-honed style
- Anyone interested in tech industry financialization
- Labor-focused readers curious about automation's real effects
What to expect
- A short, fast polemic rather than a technical deep dive
- Economic history woven through personal, workplace-level examples
- Blunt, funny prose with strong opinions upfront
- Some repetition in the book's final stretch
What carries the book past its length is voice. Doctorow writes the way he blogs: fast, funny, allergic to jargon, quick to reach for a vivid analogy over an abstract one. He's good at making a reader feel the specific indignity of being managed by software, and he's better than most tech critics at explaining financial mechanics without condescension. The result is a book you can read in an afternoon and argue about for a week, which is probably exactly the reaction it's built for.