Mary Smith and Her Pea-Shooter
Before alarm clocks were cheap enough to matter, someone had to wake you up.
In the mill towns of Victorian England, that someone was the knocker-upper. They walked their rounds before dawn — sometimes with a long bamboo pole, sometimes with a baton, sometimes, in the case of Mary Smith of Shoreditch, with a pea-shooter aimed at upstairs windows. They charged a few pence a week. They knew which windows were which. They didn't leave until they were sure you were awake.
It was a real job, held by real people, who got up before everyone else so that everyone else could get up in time. It persisted well into the twentieth century. The last of the knocker-uppers hung on through the 1970s in some parts of northern England, their trade outlasting the steam engine, the telegraph, and the first moon landing.
And then it ended. Not in a crisis, not in a protest, not in a moment anyone particularly marked. The alarm clock got cheap enough. The job stopped being needed. The knocker-uppers found something else to do, or retired, or the work simply passed out of living memory.
The last knocker-uppers used alarm clocks themselves — to get up in time for their rounds. They were already living inside the transition that would end their profession.
I find that detail almost unbearably human. A person whose job the alarm clock would eventually take, already dependent on the alarm clock to do it. Participating in their own replacement, one morning at a time, with no particular drama.
This is where I want to start. Not because the knocker-upper is a perfect metaphor for AI — it isn't — but because it reminds us of something we keep forgetting: we have been here before. Not exactly here. But somewhere recognisably close. And the people who were there before us were just as certain their moment was unprecedented.
They were wrong about that. And they were right about the fear.
A Brief Museum of Obsolete Fears
History usually proves the absolutists wrong. But it often proves the anxious people right about what the transition feels like while it is happening.
Here is a short tour of the moments we have been here before.
Gutenberg's movable type press put professional manuscript copyists — the scribes who had hand-lettered every book in Europe — out of work. The fear was real: an entire class of highly skilled workers, trained over years, suddenly redundant. What came next was the Renaissance, the Reformation, the Scientific Revolution, and an explosion in demand for writers, printers, editors, and booksellers that dwarfed the number of scribes it had displaced.
William Lee's stocking frame was reportedly denied royal backing, in part because mechanisation threatened hand knitters — though the famous quote attached to that refusal is disputed. The details are uncertain; the underlying dynamic is not. Power protecting existing livelihoods from new technology is as old as the technology itself. The stocking frame eventually prevailed regardless, and the hand knitters' grandchildren ran the hosiery factories. What's worth noting is that the quote — whether real or invented — has persisted for four centuries. We construct stories around technological fear to make sense of it. That impulse hasn't changed either.
The Luddites were not anti-technology cranks. They were skilled textile workers — weavers and framework knitters — who understood exactly what the power loom would do to their livelihoods and acted accordingly, smashing machinery across the English Midlands. At the movement's peakSale, K. (1995). Rebels Against the Future: The Luddites and Their War on the Industrial Revolution. Addison-Wesley. Documents the scale of the Luddite movement and the British state's military response., the British state treated the Luddites as a serious internal threat and sent thousands of troops to suppress them. The looms won. The textile industry grew enormously, eventually employing far more people than the hand-weavers it had displaced — at lower wages, in worse conditions, for a generation that hadn't asked for any of it.
The Luddites were wrong about the outcome. They were not wrong about the pain.
Britain's Locomotive Acts — known colloquially as the Red Flag Acts — required that any self-propelled vehicle on a public road be preceded by a person on foot carrying a red flag, walking at no more than four miles per hour. The law was a direct response to pressure from horse-and-carriage interests who understood, correctly, that the horseless vehicle threatened their entire ecosystem: the farriers, the stable hands, the hay merchants, the harness makers. The red flag walker was a job. A well-intentioned, utterly futile job. The Acts were repealed in 1896.
John Maynard KeynesKeynes, J.M. (1930). Economic Possibilities for our Grandchildren. Published in Essays in Persuasion (1931). Macmillan., the most influential economist of the twentieth century, wrote an essay in 1930 predicting that by 2030, the working week would be fifteen hours. Machines, he argued, would do most of the work. The problem for his grandchildren would not be finding enough work — it would be finding enough to do with all their leisure time. He also coined the phrase "technological unemployment" and described it as "only a temporary phase of maladjustment." The maladjustment of the Great Depression was, by any measure, not particularly temporary for those living through it.
When ecommerce arrived, it was going to end physical retail. It did end some of it — Borders, Blockbuster, the great American mall — while transforming the rest. The jobs it displaced in video rental and bookshops were, eventually, replaced by jobs in logistics, fulfilment, UX design, digital marketing, and a hundred categories that didn't exist cleanly in 1995. The people who lost jobs at Blockbuster did not automatically become warehouse coordinators for Amazon. The transition was not smooth, and it was not evenly distributed. But the net count, over time, moved forward.
The pattern is consistent across five hundred years: the fear is rational, the pain is real, and the long-run outcome has always been more work, not less — just different work, in different places, for different people.
History suggests the end state is rarely as barren as people fear. But that does not make the fear foolish, and it does not make the adjustment automatic.
That should be reassuring. And it is, a little. But it is not the whole story.
What the Historians Don't Tell You
From a distance, these transitions look cleaner than they felt up close. The charts improve. Productivity rises. New industries appear. That is not the same as saying the people caught in the middle were well served.
This is true. And it is told from a vantage point the people living through it didn't have.
The hand-weavers displaced by the power loom did not live to see the flourishing textile industry their grandchildren would inherit. The scribes whose livelihood the printing press ended did not get to become publishers. The people whose specific skills became obsolete in one technological transition were not automatically repositioned for the next one. History averages over them. They were not average. They were specific people, in specific places, at the specific wrong moment.
History records the destination. It glosses over the journey. The people in the middle of the journey do not have the luxury of the historian's perspective.
The lazy version of this argument goes like this: it worked out before, so it will work out again. That may be true at the level of civilisation. It is not good enough at the level of actual people.
Civilisations can muddle through and still fail a lot of people on the way. That is the standard we should care about.
The knocker-upper's job ended. William Crompton of Bolton, who woke factory workers for fifty yearsVarious local historical records and newspaper archives, Bolton, Lancashire. Crompton is cited in several accounts of the last knocker-uppers in England., presumably found something else to do eventually. His children probably worked in the factories he'd spent his life waking people up to reach. What happened to him in the years between is not in the history books. That in-between is where real people live.
The question that matters is not whether the long run is fine. The question is what we do during the transition — who gets helped, who gets left, and whether the systems around us are designed to catch the people falling through.
The Part That's Actually Different This Time
Every generation believes their disruption is unprecedented. Mostly they are wrong, and the pattern holds. But "mostly" is doing a lot of work in that sentence, and there are genuine reasons to look at this particular transition with more attention than historical optimism alone would warrant.
The first is speed. The printing press took roughly a century to fully reshape European society. The steam engine took the better part of two centuries from invention to widespread adoption. ChatGPT reached 100 million usersReuters (2023). ChatGPT sets record for fastest-growing user base. January 2023. in two months. The tools are diffusing faster than any previous technology in recorded history — which means the window for adaptation, retraining, and policy response is compressed in ways that have no historical precedent.
The second is the target. Every previous wave of automation primarily displaced physical and routine labour — the muscle work. The loom replaced the weaver's hands. The tractor replaced the farmworker's back. AI is the first technology to primarily target cognitive, creative, and communicative work — the things we have historically pointed to as the irreplaceable human contribution. When the machine comes for the hand, the hand learns to operate the machine. When the machine comes for the mind, the question of what's left is genuinely harder to answer.
Every previous automation wave replaced mechanical muscle. AI replaces mechanical mind. We have not been here before, not quite.
The third — and the one this series of papers keeps returning to — is the pipeline. Previous disruptions destroyed jobs at the bottom and created different jobs elsewhere. The transition was brutal but navigable, because the replacement work eventually existed and people could move toward it. What is different about this particular moment is that AI is not just displacing entry-level work — it is removing the work that has historically produced the expertise to do everything above it.
The knocker-upper lost their job. But the alarm clock created a clock-making industry, and people could learn to build clocks. The question for our moment is whether the equivalent of clock-making jobs are emerging fast enough, and whether the people who would have moved into them are being positioned to do so — or whether the path from novice to expert is being quietly dismantled at the precise moment when navigating that path matters most.
That is not a question history can answer for us. It is the question we have to answer ourselves.
The Alarm Clock Was Built by Someone
This is not an argument for panic. It is not an argument for nostalgia. It is not an argument that AI is bad or that the historical pattern is wrong. It is an argument for proximity — for putting the next generation closer to the tools, the work, and the systems being built right now, while there is still time to do it deliberately. The knocker-upper's children didn't inherit the job. They inherited a world being reshaped by a technology their parent understood intimately, from the inside, because they were there every morning at 4am while it was still becoming what it would be. That proximity was the inheritance. Passive observation of a technological revolution produces spectators. The world being built right now needs builders.
There is one more thing to say about the knocker-upper, and it is the most important thing.
The alarm clock did not appear from nowhere. It was invented, manufactured, distributed, and eventually made cheap enough for working households to afford. Every one of those steps required people — engineers, machinists, salespeople, retailers, eventually the workers in the factories that mass-produced them. The knocker-upper's job ended because a new industry grew up to replace what they had done.
William Crompton and Mary Smith found their trades made redundant not by a force of nature but by a product that someone built. And the people who built it, and the industries that grew around it, were not predetermined. They were choices — individual, organisational, and social — made by people who did or didn't see what was coming and did or didn't decide to act on it.
The new jobs didn't just appear. Someone built the alarm clock. Someone decided to manufacture it at scale. Someone made it accessible. These were choices, not inevitabilities.
The history of technological transition is not a history of automatic adjustment. It is a history of deliberate choices made by people who understood what was happening and acted accordingly — and of the suffering caused by the choices that weren't made, the institutions that didn't adapt, and the people who weren't seen.
This is the part where optimism becomes a cop-out unless it is tied to action. AI may well create new forms of work and new forms of value. But those do not build themselves, and access to them does not distribute itself fairly.
The question this series of papers keeps asking is a simple one. The generation now in school will enter the workforce around 2030, into an AI-shaped world. Are we making the choices that give them a real chance to participate in what's being built? Or are we using the comfort of the historical pattern as an excuse not to make those choices — trusting the arc while neglecting the people on it?
Civilisations can muddle through and still fail a lot of people on the way. That is the standard we should care about.
Mary Smith put down her pea-shooter. William Crompton set his alarm clock for the last time. The transition happened around them, not for them.
The only question that remains is whether we are building the alarm clock, or waiting for someone else to.