History rhymes#

Over the past few months, we’ve been witnessing a genuine revolution driven by AI. Tasks that until recently required skilled humans can now be performed by AI in a fraction of the time and cost. This white-collar revolution isn’t without echoes of the blue-collar revolution that transformed the world centuries ago.

The industrial revolution#

Starting around 1760, machines began replacing skilled craftsmen—weavers and spinners who had spent years mastering their trade.

Early machines actually produced inferior goods—spinning jenny thread was weaker than hand-spun yarn. But machines improved rapidly, and within two decades were producing thread that was finer, stronger, and more uniform than anything a human could make. Consistency became the machine’s greatest advantage: handcraft varied by craftsman, but machines delivered predictable results every time.

The speed differential was brutal. A single operator could manage over a thousand spindles where a craftsman had managed one. The economic logic was inescapable. Prices dropped 60-70%. Demand exploded. What once required master craftsmen could now be done by semi-skilled operators.

The transition took 60-80 years to stabilize. Brutal as it was, multiple generations had time to adapt—children grew up in a world their parents had learned to navigate.

The AI revolution#

Now consider what’s happening with AI.

The pattern looks familiar. Early AI output was often worse than human work—hallucinations, errors, confidently wrong answers. But improvement has been breathtakingly fast. In late 2024, we had basic code completion. By early 2026, we have autonomous agent swarms tackling complex tasks. The quality gap is closing at a pace the Industrial Revolution never saw.

The speed advantage is comparable or greater. What took a developer hours can now take minutes. What took days can take hours. And unlike the spinning jenny, AI doesn’t just do one task faster—it accelerates across an enormous range of knowledge work: writing, coding, analysis, design, research.

Costs are dropping. Demand is exploding. Knowledge work that once required expensive experts is becoming accessible to anyone with a subscription. The democratization is real.

But the revolutions diverge on one crucial point: the Industrial Revolution took 60-80 years to stabilize. The AI revolution is moving perhaps a thousand times faster. We’re not talking about generations of adjustment—we’re talking about years, maybe months.

And the breadth is staggering. The Industrial Revolution primarily affected manual labor in specific industries—textiles, manufacturing, agriculture. The AI revolution touches nearly every white-collar profession simultaneously: software, law, medicine, finance, design, writing, research. There’s nowhere to pivot to.

The return to craftsmanship#

Two centuries after machines won, we’re seeing a quiet revival. Craft beer, handmade furniture, bespoke clothing—people pay premium prices for human touch, even when machine-made is technically superior. Etsy built a billion-dollar business on it.

But the real test came when Notre Dame burned in 2019.

To rebuild the cathedral, France needed stonemasons who understood centuries-old methods, stained glass artisans, master roofers for lead work. They had to recruit 2,000 skilled workers, including 60 master carpenters from around the world.

Machines couldn’t do the work. After 850 years of settlement, nothing in the cathedral was level or straight. The craftsmen had to “reproduce all of the deformations that had accumulated over eight centuries.” No algorithm could handle that. The skills had nearly died—and when they were needed, the world had to scramble.

Minds, not hands#

The Industrial Revolution threatened livelihoods. The AI revolution may threaten something deeper.

Manual skills took humanity hundreds, perhaps thousands of years to develop. But thinking—the capacity for reason, abstraction, creativity—took millions of years of evolution. And unlike a muscle that can rest, the mind requires constant stimulation to stay sharp. That’s precisely what AI threatens to take away—something I’ve explored in how developers learn and its AI sequel.

There’s another difference. Much of our manual knowledge could be written down, preserved, taught. Our intellectual potential is harder to capture. It’s not just what we know—it’s our capacity to generate new knowledge. That’s not so easily stored.

And here’s the troubling part: AI needs material to learn from. It has already absorbed virtually everything humanity has ever produced. Now it increasingly trains on content it generates itself—a kind of intellectual inbreeding whose outcome we can predict all too well.

The first revolution spanned generations. This one is moving in months. It’s faster, broader, and more intimate—it touches not our hands but our minds.

The question is no longer just economic. It may be existential.

The question isn’t whether AI will transform knowledge work—it already is. The question is whether we’ll preserve the deep expertise we might desperately need later.

Notre Dame needed craftsmen no machine could replace. Someday, we may (dare I say will?) face problems that need thinking no AI can replicate—work that requires feeling the problem, adapting to its irregularities, leaving human traces in the solution.

Will those craftsmen still exist? Or will we discover, too late, that the skills have died?

That’s the Notre Dame question. And unlike the cathedral, we might not get five years to find the answer.