The best paid work of the last century happened sitting down. Nobody announced this as a policy; it accumulated. Jobs moved indoors, then onto screens, and somewhere along the way we sorted work into a quiet hierarchy where distance from physical labor tracked status almost perfectly. The body became a thing you maintained on evenings and weekends so it could carry your head to a desk.
Then automation arrived in the wrong order. For decades the standard forecast was that machines would climb the ladder from the bottom: warehouse first, truck driver next, the office jobs last and maybe never. Instead the models learned to write, code, and analyze, while the warehouse stayed stubbornly full of people. The body we spent a hundred years devaluing is turning out to be the harder engineering problem, and I think the consequences of that inversion go beyond a labor market footnote.
Moravec saw it coming
This wasn't unforeseeable. The roboticist Hans Moravec pointed out back in the 1980s that we had the difficulty rankings backwards: abstract reasoning is evolutionarily recent and comparatively easy to engineer, while perception and movement, the things a toddler does without effort, rest on hundreds of millions of years of optimization and are brutally hard to replicate. We judged skills by how hard they feel from the inside, and walking across a cluttered room feels like nothing.
If you want the vivid version, watch footage from the 2015 DARPA Robotics Challenge, where robots built by some of the best teams in the world were asked to open doors, turn valves, and walk across rubble, and fell over. Repeatedly, expensively, in slow motion. In that same era software was already far beyond any human at chess and about to be superhuman at Go. Robotics has improved a great deal since, but a decade later there is still no machine you would trust to fix the plumbing in an unfamiliar house.
Knowledge that never made it into words
The plumber matters here, actually. When someone who has spent twenty years with their hands on pipes walks into your flooded kitchen, touches the fitting, listens for a second, and tells you what's wrong, you are watching a real form of intelligence, built out of thousands of unrepeatable physical situations. None of it was ever written down. That matters more than it used to, because the current generation of AI learned from what got written down. A surgeon's feel for tissue, a mechanic's ear for a failing bearing, the way a good nurse handles a frightened patient: same category. This knowledge sits outside the training data, not because it's secret but because language can't hold it.
You can already see some repricing. Skilled trades have been getting harder to hire across most rich countries for years, and the pandemic gave everyone a blunt preview of which work counted as 'essential' when everything closed: nurses, drivers, shelf stockers, people whose jobs happen in a body, in a place.
What I'm not claiming
Not fitness culture. The optimized, tracked, photographed body is mostly the mind's project imposed on flesh, one more dashboard to manage. And not a back-to-the-woods romance either; I build software for a living and I like it. The claim is narrower and drier: the price of embodied skill is being pushed up by the simple fact that our most powerful tools got good at everything else first.
The part I'm least sure about
The honest objection is that the moat could be temporary. Humanoid robots attract serious money now, warehouse automation keeps improving, and I would not bet much on fine manipulation in messy environments staying unsolved for thirty years. Maybe it falls in fifteen. I don't know, and people who claim to know are guessing.
But half of what I'm describing was never about capability. Call it presence: being fully in a room with another person, undivided. When something terrible happens to someone, the friend who physically shows up does something no message can, and not because their body performs a task. The visit means what it means because a finite person spent a piece of their finite time being there. A machine can eventually simulate the sitting. It can't spend anything.
There's an irony I might as well admit: I help build an assessment platform, and everything it measures has to pass through a screen. The kinds of intelligence in this essay are exactly the ones it can't see. Most of our systems for deciding who is capable share that blind spot, which is probably part of how embodied skill got undervalued in the first place. If it can't be tested online, it's invisible to the sorting machinery.
So I've been trying, without much consistency, to treat presence as a skill that needs practice rather than a mood that shows up on its own. Attention drifts to the device by default now; being where you are takes deliberate effort, which is a strange thing to have to say about standing in a room. I'm not sure how far the economic reversal runs. The personal one seems worth making regardless.