Progress Studies · The Diffusion of Technology
For nearly all of human history, almost nothing changed. Then, in the blink of a few centuries, everything did. This is the story of the inventions that bent the curve — and the strange, stubborn lag between having an idea and living in the world it makes.
The shape of the whole story
A person born in the year 1000 lived much like their great-great-grandparents: the same tools, the same crops, the same short life. Measured by what an ordinary person could buy, eat, or expect to survive, the world barely moved for millennia.
Then the line breaks upward so sharply it looks like an error in the data. It isn't. It's the compounding effect of a handful of general-purpose technologies — each one a new foundation the next was built on top of.
The hockey stick of human prosperity
World GDP per person · 1990 international $ (Maddison) · illustrative
The idea behind this page
"Progress studies" is a young field — named in a 2019 essay by Patrick Collison and Tyler Cowen — that asks an unfashionable question: what actually causes progress, and could we get more of it, faster?
One of its most useful findings is also the most counterintuitive: a breakthrough almost never pays off when it's invented. It pays off decades later, once the world reorganizes itself around it.
The dynamo took forty years to show up in the productivity statistics.
Factories electrified slowly not because the technology was missing, but because a steam-era factory was built around a single giant engine and a spider's web of drive shafts. Electricity only paid off once someone thought to give every machine its own small motor and lay the factory out around the work instead of the power source. The economist Paul David used this to explain a puzzle of his own era — Robert Solow's 1987 quip that "you can see the computer age everywhere but in the productivity statistics." The computers did eventually show up in the numbers. It just took until the late 1990s.
Everett Rogers gave this lag its canonical shape in 1962: the S-curve of diffusion. New technology creeps along among a few innovators, hits a tipping point, races through the mainstream, then flattens as it saturates. The interesting action — and the impatience — lives in the slow early years before the curve turns.
Descend through time
Each marker notes how long since the previous one. Watch the intervals collapse — from millennia, to centuries, to years, to months.
Farming turned wandering bands into settlements with a food surplus — and a surplus is what lets some people do something other than find dinner: build, rule, write, invent.
Diffusion: the slowest in history. Domestication spread across continents over thousands of years, arising independently in at least seven separate regions.
In Mesopotamia, marks in wet clay externalized human memory. Contracts, laws, and knowledge could now outlive the person who held them — the substrate all later progress is recorded on.
Diffusion: writing existed for over 5,000 years before most people could read it. Literacy stayed below 10% of humanity until the 1800s.
Gutenberg's movable type collapsed the cost of copying an idea by orders of magnitude. Knowledge stopped being a scarce, hand-copied luxury and became something that could go viral.
Diffusion: fast for its age — presses reached 200+ European cities within ~50 years. But its full payoff (the Reformation, the Scientific Revolution, mass literacy) unfolded over the next three centuries.
Newcomen's engine, and Watt's far better one in 1776, gave humanity power that didn't tire, didn't need feeding, and didn't depend on wind or a river. This is the machine that broke the flat line on the chart above.
Diffusion: roughly 70 years from Newcomen's first pump to steam becoming the engine of industry — factories, railways, and ships in turn.
Jenner's smallpox vaccine, then Pasteur and Lister's germ theory, revealed an invisible enemy — and how to beat it. More than any drug, clean water and vaccination are why a child born today is likely to reach old age.
Diffusion: decades to reach doctors, longer to reach everyone. The collapse in child mortality — from ~40% to under 4% — is mostly a 20th-century story.
Edison's first power station lit lower Manhattan. Light, then motors, then everything: the most general of all general-purpose technologies, a clean force you could pipe into any building and do anything with.
Diffusion: ~40 years to reach half of U.S. homes — and about as long for factories to redesign themselves enough to see the productivity gains. The textbook case of the lag.
The Model T made the car a thing ordinary people owned, not a rich man's toy. It redrew the map — suburbs, highways, supermarkets, the whole geography of modern life follows from it.
Diffusion: ~30 years from the Model T to mass ownership, accelerated by Ford's assembly line dropping the price faster than any product before it.
Fleming's penicillin made once-fatal infections trivial. Decades later, Norman Borlaug's high-yield wheat let farmers feed billions more from the same land — a quiet achievement often credited with averting mass famine.
Diffusion: penicillin went from lab to battlefield in ~15 years, spurred by WWII. Borlaug's seeds spread across Asia in a single decade.
The transistor (1947) shrank into the microprocessor: a computer on a sliver of silicon. Moore's Law then doubled its power every couple of years for half a century — the steepest sustained improvement of anything, ever.
Diffusion: here Solow's paradox bites — computers were "everywhere except the productivity statistics" through the 1980s. The payoff finally landed in the late 1990s, ~25 years on.
ARPANET (1969) connected the machines; Tim Berners-Lee's World Wide Web (1991) made the connection usable by anyone. Distance stopped mattering for information — the cost of sharing anything with everyone fell to nearly zero.
Diffusion: ~7 years from the first web browser to mainstream — one of the fastest adoptions on record, until the next two.
The iPhone folded the computer, the internet, the camera, the map, and the wallet into one object in everyone's pocket. For most of humanity it became the primary way they touch the digital world at all.
Diffusion: ~5 years to ubiquity in rich countries, and it leapfrogged straight past PCs across much of the developing world.
ChatGPT put a system that reads, writes, codes, and reasons in front of anyone with a browser. For the first time, the thing being automated isn't muscle or arithmetic — it's cognition itself.
Diffusion: ~2 months to 100 million users — the fastest-adopted consumer technology in history. The open question is the oldest one on this page: how long until it shows up in the productivity statistics?
The acceleration, in one chart
The gap between invention and impact used to be measured in generations. Adoption — how fast a new thing reaches a hundred million people — has been collapsing just as dramatically. The telephone took three-quarters of a century. ChatGPT took a long weekend.
How long to reach 100 million users
years from launch · longer bars = slower to spread
So where does AI fit?
The whole point of the lag is that you can't feel it while you're inside it. In 1890, electricity was a novelty for lighting; the factory revolution it would cause was still thirty years out and invisible to almost everyone.
That's the optimist's case for AI: the models exist, but the world hasn't yet reorganized around them — the equivalent of rewiring every factory floor is still ahead. If the pattern holds, the real gains arrive not when the technology is announced, but when institutions, jobs, and habits quietly rebuild themselves around it. That takes years, and it's rarely the incumbents who manage it.
The pessimist's case is that this time the lag could be short in a way that's its own kind of problem — that cognition spreads faster than society can adapt to it. Both camps are staring at the same chart. Progress studies is the discipline of trying to read it before the ink is dry.
The line on the first chart is still climbing. What bends it next — and how long we wait to feel it — is the question this whole field exists to ask.