The argument
The productivity gains from artificial intelligence are flowing to capital, not to labour, and that is not a law of physics. It is a design choice, renewed every quarter by institutions we have not bothered to update.
This book refuses both comforting stories: it is not a warning about robots, and it is not a manifesto for a workless utopia funded by a universal cheque. Redistribution taxes the winners after the fact and leaves ownership exactly where it is. The alternative is predistribution: a citizen’s stake in the machines, built before the gains are gone.
I have spent thirty years building the software that lets companies produce more without hiring more. This is the view from inside the machine.
Writes better English than many English speakers, and is a better economist than the ‘real’ economists.Martin Schneider, the book’s copy-editor
Chapter one — Who stole the gains?
It is midnight in Changping, on the northern outskirts of Beijing, and the factory floor is running at full capacity. Robotic arms sweep through precise arcs, their movements plotted to the millimetre by algorithms that never tire and never deviate. The production line is assembling premium smartphones at a rate of roughly one every three seconds: faster than you can read this sentence aloud.
No one is there to see it.
The lights are off.
The chapter continues in the book: through the dark factories, the fifty-year split between pay and productivity and the question the technical fixes never reach.
The short version, free
The same argument, the same facts, the same chapters, in a third of the length and in plain words. For readers who want to understand what the machines do to work, wages and ownership, without the economics. A free epub in English, Dutch and French, out 30 September.
A guide to the chapters
What each chapter argues, in plain language, on a single page you can print and hand on. Written for people who are reading the book and do not follow the economics, and free in English, Dutch and French.
The book in six conversations
Two AI hosts discuss the book about who owns the machines, deliberately. A six-episode series generated from the manuscript and fact-checked against it, one episode a week.
The Stable: the book’s claims, scored against the news
September 2026
The empty lecture halls
The University of Antwerp opened the academic year with 29 per cent fewer first-year computer science students than last year, after a fall of almost 35 per cent the year before. The city's college lost 40 per cent of its new programming students. Both name the same suspect: young people believe AI will do the programming. Chapter 3 predicted a squeeze at the entry rung from US payroll data. Belgium's High Council for Employment expects it, Dutch payroll data show it next door, and in September the supply side answered. The people who would have been the next juniors are not enrolling.
September 2026
Slow down, sell up
On Saturday 12 September Dario Amodei wrote that the industry must slow the pace at which it improves AI models, and Sam Altman, Elon Musk and Demis Hassabis agreed in public. The same day Altman told Fortune that OpenAI will not list in 2026. The same weekend Anthropic picked Nasdaq for an October listing that reports put near 2 trillion dollars, twice the 965 billion at which its private round valued it in May. The warning and the money moved in the same week and in opposite directions. On the register the warning changed nothing: the rise was banked in private and the public is offered the top, while OpenAI's asset stays private outright.
The author
David Vanheeswijck builds AI systems for a living: co-founder of Hiveras, an AI collective designing autonomous agents for businesses, and by day a principal engineer on the systems behind customs and trade. Not an economist, but a practitioner who watches the automated economy arrive inside his own tools, week by week.