Why We Should Be Wary of ‘Clean’ History
In a recent exchange about the current state of AI—and the strange tension between ‘slop’ output and ‘human-level’ reasoning—the conversation inevitably turned to the Industrial Revolution. It usually does. When we are faced with a technology that feels like it might be a ‘clean break’ from the past, we go looking for historical precedents to help us build a narrative for the future.
The problem is that we often go looking for clean mechanisms in the past to justify our predictions for the present.
The Allure of the ‘Clean’ Narrative
There is a school of economic history that works backward from a preferred answer. They start with a thesis—say, ‘The Enlightenment created a culture of growth’—and then marshal evidence to support it.
Joel Mokyr is perhaps the exemplar here. While his work is essential reading, he tends to start with a thesis and walk back the causality only when forced. Gregory Clark is another; his confidence often masks a certain circularity. These ‘clean’ histories are popular because they make the world feel legible. They suggest that if you just get the ‘inputs’ right—the right culture, the right institutions, the right data centers—progress is an inevitable output.
But if you appreciated David Landes for his willingness to sit with complexity rather than impose a mechanism, you quickly realize that the “clean” versions of history miss the most important part: the texture.
A Syllabus for the Messy Reality
If we want to understand what technological transformation actually looks like when it hits a complex human system, we need to look at the ‘multi-causal messiness.’ Here are the works I suggested for when you want to see the gears of history without the social-scientific smoothing:
- The Materialist Correction: Robert Allen, The British Industrial Revolution in Global Perspective. This is Mokyr done ‘honestly.’ Allen starts from observable material conditions—expensive labor and cheap coal—and lets the mechanism emerge from that. It’s an ‘induced innovation’ framework that is actually falsifiable.
- The Human Texture: E.P. Thompson, The Making of the English Working Class. A completely different political register from Landes, but with the same commitment to letting the past be strange. It captures the sheer friction of people being forced into new ways of living.
- The Coercive Engine: Sven Beckert, Empire of Cotton. Beckert refuses to let you separate industrial dynamism from coercion. It reminds us that ‘progress’ isn’t just a series of clever inventions; it’s often a reorganization of power and violence.
- The Systemic Sceptic: Charles Kindleberger, Manias, Panics and Crashes. Essential for anyone watching the AI markets today. Kindleberger is deeply skeptical of clean economic mechanisms and brilliant on how human behaviour actually unfolds during ‘revolutions.’
- The Long View: Fernand Braudel, The Wheels of Commerce. Braudel’s longue durée approach is the opposite of a monocausal narrative. It rewards patience and reveals how deep-seated systems of trade existed long before the ‘big bang’ of industrialization.
Why Texture Matters Now
We see ‘AGI on Mars‘ because we want a clean narrative of progress. We want to believe that ‘reasoning tokens’ lead directly to ‘A’ and then ‘B.’
But as Landes showed in his earlier work, like Bankers and Pashas, progress is always snagged on the human element. It is messy, it is contested, and it rarely follows a straight line. If we strip away the texture of the past to find a clean mechanism, we end up building a map for a world that doesn’t exist.
The current ‘Progress Studies’ movement is looking for a master key to growth. But the historians who sit with the texture of the past suggest there is no key—only a massive, multi-causal, and often contradictory tangle of events. If we want to build the future, we should spend less time looking for a clean mechanism and more time learning to navigate the mess.
The value isn’t in finding the ‘answer’ to the Industrial Revolution. The value is in developing a tolerance for the mess.