In a context where AI dominates our conversations about the future, Carl Benedikt Frey reminds us of an uncomfortable truth: technological progress is neither automatic nor guaranteed. For most of human history, stagnation was the norm, and even today, progress in the world’s largest economies — the United States and China — has not met expectations. This reflection comes at a timely moment as we seek to understand why some societies thrive while others fail in the face of rapid technological change.
The Economist-Historian Who Questions Our Certainties
Carl Benedikt Frey is a Swedish-German economist and economic historian, Associate Professor of AI and Work at the Oxford Internet Institute and a fellow of Mansfield College at the University of Oxford. After studying economics, history, and management at Lund University, Frey completed his doctorate at the Max Planck Institute for Innovation and Competition in 2011 and founded the Oxford program on the future of work.
It was in 2013 that Frey, with his colleague Michael Osborne, co-authored “The Future of Employment: How Susceptible Are Jobs to Computerization,” a study that, with over 20,000 citations according to Google Scholar, employed a methodology adopted by President Barack Obama’s Council of Economic Advisers, the Bank of England, and the World Bank. This experience with technological forecasting naturally led him to question the deep mechanisms governing progress and its cessation.
The Innovative Thesis: The Precarious Balance of Progress
Frey’s work examines technological and economic development over the last 1,000 years, arguing that it rests on a balance between decentralized innovation and centralized implementation. Frey uncovers a recurring tension in history: while decentralization fosters the exploration of new technologies, bureaucracy is crucial for deploying them at scale. When institutions fail to adapt to technological change, stagnation inevitably follows. Only a careful balance between decentralization and bureaucracy enables nations to innovate and grow over the long term.
This vision radically nuances our understanding of progress. “Progress depends on a delicate balance between decentralized innovation and bureaucracy for deployment,” as Frey summarizes in his presentation of the work. Centralized bureaucracies excel at speed — exploiting established technologies and deploying them rapidly — while decentralized systems are better suited to experimentation.
Innovation Flourishes in the Margins, Not in Ministries
Frey insists that progress depends on dynamic efficiency. He admires the capacity to generate and test new ideas rather than the efficiency of existing production. Innovation often emerges from ideas that blossom in informal social networks — English coffeehouses or American salons.
The Industrial Transformation was, according to Joel Mokyr’s well-known argument, characterized by inventors and tinkerers who experimented with new technologies and adapted them alongside others in learned societies and informal networks. This is precisely the horizontal diffusion of knowledge that Frey considers key to decentralized innovation. As Frey notes, the “Birmingham Lunar Society” included the inventors of latent heat (Joseph Black) and the steam engine (James Watt), and the Industrial Transformation took place in Britain because of a lack of “stifling” centralized bureaucracy.
Breakthroughs like mRNA vaccines stem from decades of decentralized and uncertain experimentation. But beware of myths: Frey points to Silicon Valley itself as a product of this decentralized dynamism, yet here the narrative risks becoming mythologized: Google, often celebrated as the darling of scrappy startup culture, succeeded not through pure garage tinkering but by acquiring competitors with substantial venture capital support from Sequoia and Kleiner Perkins before going public.
Bureaucracy: Necessary Evil or Hidden Engine?
Frey’s later chapters trace the rise and limits of bureaucratic planning. Both capitalist and communist states achieved rapid postwar growth through centralized coordination, but these efforts stagnated because rigidified institutions failed to adapt to technological change — particularly in the USSR, which could not reproduce the decentralized innovation of Silicon Valley.
The Chinese example fascinates Frey: China occupies a complex position in Frey’s constellation. It possesses the most historically established state bureaucracy, established over 1,000 years ago. Counterintuitively, according to Frey, China’s massive meritocratic state stimulated technological development from 1000 to 1400, making it the most technologically advanced political entity in the world.
According to Frey’s argument, Meiji Japan and Bismarckian Germany succeeded only because technological innovation had occurred elsewhere (through decentralized, horizontally structured organizations). Two key examples of decentralized “exploration” are the Industrial Transformation in the United Kingdom (roughly 1750 to 1900) and the so-called Second Industrial Transformation in the United States (roughly 1870 to 1914).
The Ignored Lessons of Stagnation
Frey examines why certain past technological powers — such as Song China, the Dutch Republic, and Victorian Britain — ultimately lost their innovative edge, why some modern nations like Japan experienced periods of rapid growth followed by stagnation, and why planned economies like the Soviet Union collapsed after brief bursts of progress.
Russia under Peter the Great was a centralized state that was too autocratic to implement new technologies at scale. The Soviet Union was able to catch up in the mid-twentieth century through vertical hierarchical organization and bureaucracy. Though this centralized approach worked for heavy industry, it did not work for the information age, and their failure to capitalize contributed to the decline in growth during the 1970s and 1980s.
By the 1980s, increasing bureaucratic centralization had failed across East and West: incumbent firms stifled any disruption to their dominance. Globalization and digital technologies concentrated innovation in a few large companies, leading to a slowdown in productivity and creativity — what Frey calls “The Great Flattening.” He also explores progress and authoritarian regression, contrasting China’s AI-driven surveillance state with that of the USSR. He suggests that while authoritarianism can bring short-term technological gains, the kind of iconoclastic innovation necessary for long-term progress is diminished.
The Warning for the AI Era
Frey worries that both China and the United States face this challenge: China because of the centralization of power under Xi Jinping, and the United States because incumbent firms have succeeded in stifling competition. The book warns that both the United States and China are heading toward the same type of stagnation. Vested interests are hampering the United States’ ability to harness its decentralized economy for innovation, while increased centralization in China jeopardizes their dynamism.
This vision contrasts with prevailing technological optimism. The book concludes that conditions in the United States and China today point toward a future of stagnation despite their dominance at the AI frontier, because major inventions from the steam engine to generative AI disrupt societies before they bring improvements in living standards.
The Shadows in the Analysis
Frey demonstrates remarkable mastery of a millennium of economic history, but his theoretical framework raises questions. Frey’s main idea is that distributed power drives more innovation, while centralized power is better for exploiting the results of that innovation. The book is primarily a collection of historical anecdotes supporting this thesis, and though it seemed there would be enough of them, the stories were sufficiently interesting in themselves.
The role played by geopolitical rivalries is one that should have received far more discussion, given the examples chosen in the book, but the desire to conclude toward a predetermined set of ideas likely prevented the author from attempting to explain why factors that hindered the USSR, for instance, might not matter for China.
The analysis sometimes remains too binary in its decentralization/centralization vision, neglecting the hybrid forms emerging today. Moreover, Frey perhaps devotes too much attention to institutional structures at the expense of cultural and social factors that shape innovation.
Why Read This Book Now
What struck me most was the book’s analysis of AI and its limitations, and especially the point that even in the AI age, you still need a distributed network of research and enterprise to achieve genuine progress. This is something that truly warrants reflection.
What makes Frey’s ambitious itinerary so compelling is that it reframes inherited wisdoms in light of today’s dilemmas and insists that progress can only be understood when the past is brought into dialogue with the present.
This book is for anyone questioning technological futures without resorting to prophecy. Frey brings a historical perspective rare in a debate dominated by exponential extrapolations. For policymakers, entrepreneurs, and citizens concerned about AI’s impact, the work offers a robust intellectual framework for understanding why progress stops — and how to avoid it.
‘How Progress Ends’ is ultimately not a pessimistic book but a disciplining one. It asks us to abandon our illusions of inevitability and instead direct our attention to the fragile mechanics of innovation — networks, institutions, and cultures — that shape human futures.
Bibliographic Information
Title: How Progress Ends: Technology, Innovation, and the Fate of Nations
Author: Carl Benedikt Frey
Publisher: Princeton University Press
Publication Date: September 2025
Pages: 552