In 1882, Edison switched on the Pearl Street Station and electricity became something a factory owner could buy. The dynamo worked. Owners bought in. And then, for about forty years, almost nothing happened.
That is the part everyone skips. The technology arrived and the productivity did not. As late as 1899, electric motors drove under five percent of American factory machinery, and the broad measurable payoff did not land until the 1920s, the better part of four decades after the dynamo went on sale. The economist Paul David documented this in 1990, using the lag to explain why the computers of his own decade were not showing up in the productivity statistics either.
The machine was never the bottleneck. The arrangement of the building was.
The factory had to be rebuilt, not re-engined
#An early electrified factory was a steam factory with a different engine bolted on. A steam plant runs one giant central driveshaft, and every machine in the building hangs off that shaft by a belt. The whole floor plan is organized around the shaft, not around the work. So owners pulled out the steam engine, wired in one big electric motor to spin the same shaft, and changed nothing else. They had electrified the power source and kept the factory exactly as dumb as it was before.
The gains came later, when someone worked out that you could put a small motor on every machine. Unit drive. Now a machine no longer had to sit within belt’s reach of the shaft. You could lay the floor out around the flow of the work, build on one level, light it sanely, move materials in a straight line. The motor did not deliver the productivity. The redesign did, and the redesign took a generation of people figuring out what the motor had quietly made possible.
History keeps making this point with different props
#Consider the English longbow. The weapon itself was a stick and a string, cheap to make and simple to understand. The thing that was neither cheap nor simple was the archer.
A war bow pulled somewhere between 100 and 160 pounds, several times a modern hunting bow, and you did not develop the body to draw one as an adult hobby. You did it from childhood. We know this fairly literally: the skeletons of the archers recovered from Henry VIII’s warship the Mary Rose carry the marks of a lifetime of it, including an unusually high rate of a specific shoulder adaptation that comes from drawing a heavy bow before the bones finish fusing. England wrote the training into law. A 1541 statute required that boys be given a bow from the age of seven and “brought up in shooting,” and earlier ordinances mandated practice and banned the games that competed with it.
That is the actual edge, and it is worth being precise about what it was not. It was not that the arrows punched through plate armor, because they mostly did not: a knight in fitted fifteenth-century plate was close to arrow-proof, and the longbow never won a battle single-handed, mud and stakes and terrain and the other side’s mistakes always did their share. The edge was that England had manufactured, at national scale and over years, a deep bench of people who could do a hard thing on demand. Its rivals reached for the crossbow precisely because you could train a crossbowman in a couple of weeks. The crossbow put the power in the mechanism. The longbow put it in the human, which is slower to build and much harder to copy.
The pattern has a name
#Two technologies, five centuries apart, the same shape: the tool shows up fast, the human and organizational adaptation shows up slow, and the slow part decides who comes out ahead. The adaptation gap is the lag between what a technology makes possible and how fast people reorganize themselves to use it. It is the real story of every major shift, and the machine is the easy half.
Which brings us to the mood
#The disruption coming from AI is real, and I am not going to try to talk you out of it. A lot of work is going to change, and a lot of it faster than anyone is comfortable with. Pretending otherwise is its own kind of snake oil.
The public has clearly priced in the fear. In the most-cited recent read on this, only 17 percent of Americans said AI would have a positive impact on the country over the next twenty years. Among the AI experts surveyed, that figure was 56 percent. You can read the 39-point gap two ways. One is that the experts are high on their own supply. The other is that the people closest to the thing can see a payoff the headlines are not pricing in. In my experience both are a little true at once, which is the least satisfying kind of answer and usually the correct one.
But the dynamo and the longbow point at the same thing. The outcome was never set by the machine. It was set by how fast the humans adapted around it, and unlike the technology, that part is a choice.
So point it at people
#The useful question was never whether AI is good or bad for people in the abstract. It is where you aim it.
You can aim an acceleration at replacing people. That is the default, it is the cheap option, and it is the modern version of bolting a motor onto the old shaft and declaring the factory upgraded. Or you can aim it at making people more capable, and faster at becoming so, which is the harder rearrangement and the one with the real gains buried in it.
You can see the fork on a real team. Point the tool at the work and you get a thinner org doing roughly the same things slightly cheaper, and a memo. Point it at the people and the junior analyst who used to lose three days a month pulling numbers spends them on the judgment call instead, and the line manager who was quietly sure the motor on the wall was coming for her job turns out to be the one person who knows what to do with it. Same technology, opposite outcomes. The only variable was where it got aimed.
We have run a version of this experiment before and written down the result. For years after a computer first beat the world chess champion, the best entity in chess was not the strongest computer. It was an average human paired with a machine, what Garry Kasparov started calling a centaur. The lesson was never that the machine won. It was that the pairing won, and it took people a while to believe it.
The optimistic part, and I will admit I spent a PhD looking at how the brain reorganizes itself, is that people adapt far better than we give them credit for. The wetware is built for exactly this. What we are bad at is not the learning. It is rebuilding the factory around the learning, which is an organizational failure we keep blaming on individuals.
The motor has been on the wall for a few years now. Most of us are still standing around admiring it, waiting for it to do the work. It is not going to. Pointing the acceleration at people instead of around them is the harder rearrangement, the one with the gains buried in it, and for lack of a better word I have started calling it human acceleration. We are the part that has to move.