The New Laws of Motion
It took three sentences to explain every moving thing in the universe. The trick was imagining away friction. The economy has just met its first frictionless actor — and the laws are coming into view.
In 1687, a man who had spent years grinding his own lenses and feuding with half of European science published three sentences that explained every moving thing in the universe.
This was someone who had once slid a blunt needle into his own eye socket, behind the eyeball, to see how pressure changed colour. He wrote it up in his notebook with the calm of a man describing the weather.
Cannonballs, comets, tides, the Moon. Newton’s Principia reduced all of it to three laws you can write on the back of a hand.
The trick wasn’t better observation. Humans had been watching things move for ten thousand years and had concluded, reasonably, that moving things slow down and stop. Aristotle built a whole physics on that, and it matched everything anyone had ever seen.
What Newton did, following Galileo, was imagine a world without friction. Galileo rolled bronze balls down inclined planes and asked what would happen if the surface were perfectly smooth. Newton took the question the rest of the way. Remove friction, in thought, since you cannot remove it in fact, and the hidden laws appear. Objects in motion stay in motion. They always did. You just couldn’t see it, because everything you had ever watched was dragging.
Economics never produced its Principia, and I don’t think the reason is a shortage of genius. The reason is that economics never had a frictionless body to study. Every economic actor in history has been a human being. We tire, we forget, we mistrust each other, we can’t read twenty pages of an insurance contract without our soul leaving our body. Strip the friction out of economics and there is almost nothing left to look at.
Then, somewhere in the last two years, the frictionless body arrived. It is not superintelligent, which turns out not to matter. It is frictionless.
And the laws are becoming visible. I think there are three.
Law I: Every toll persists until acted on by an outside force.
In 1937 a 26-year-old economist named Ronald Coase published a paper asking a question so obvious nobody had thought to ask it: why do firms exist?
If markets are so efficient, why isn’t everyone a freelancer, contracting every task at the market price? Why do we instead gather thousands of people inside organisations, with managers and memos and meetings?
Coase’s answer was transaction costs. Using the market is expensive. You have to find people, negotiate with them, check their work. Below a certain threshold it’s cheaper to coordinate humans inside a firm than to contract them outside one, and that threshold is where the boundary of the firm sits.
Read forward, the way it’s taught, it’s a theory of why companies exist. Read backward, it’s a warning. The firm is a machine for paying a tax, and if the tax ever fell to zero the machine would have no reason to exist.
The hedge fund manager Alap Shah (you might have read the essay he wrote with Citrini a few months back), on a recent episode of Hidden Forces with Demetri Kofinas, put an image on that tax. Think of the entire Microsoft Office suite, he says — Outlook, Word, PowerPoint — as one large coordination suite for humans, all of it dedicated to making people understand why and how they’re doing something. The meeting exists for the same reason. So does the manager, and most of the org chart. Take away the work itself and what remains, which is most of the payroll, is apparatus for getting more than a few humans rowing in the same direction.
Those of us who work in offices now how many hours daily are spent in coordination meetings.
Agents need none of it. They run around the clock, take no benefits, and they sing from the same context, as Shah puts it: one learns something and the rest know it in real time, no meeting required. His cost estimate is under one per cent per unit of work.
So substitution never needed superintelligence. The wrong comparison is the model against your best analyst. The right comparison is the model against the full cost of employing, aligning and managing a team of analysts, and almost-as-good clears that bar years before better does. The saving was never just the worker. It was the machinery around the worker.
Once you see one toll, you see them everywhere.
The consumer internet is a toll. Shah’s framing is that the trillions of market cap built on it monetise human time, human laziness, and human imperfect information. You don’t run three more searches to save a pound. You don’t read all the reviews, and you certainly don’t read the insurance contract; you default to the trusted app and renew. The aggregators sitting between you and the pizza, the ride, the hotel, the loan have a moat, but the moat is not the pizza. The moat is that you won’t do the comparison yourself.
Entry-level employment is a toll, paid by the firm this time. A graduate is carried at a loss for five or ten years, a training tax the organisation absorbs to manufacture the capable thirty-five-year-old it actually wants.
Even the tax code is a toll, in the literal direction. A human’s work carries seven to eight per cent in payroll taxes that an agent’s identical work does not. We are subsidising the substitution.
Why did these tolls survive for decades in markets that are supposed to compete everything away? Inertia. A body at rest stays at rest until acted on by an outside force, and there was no outside. Everyone paid these tolls because everyone was human, and no one can opt out of being human. A toll that everyone pays doesn’t look like a toll. It looks like the economy.
The outside force has now arrived. It is infinitely patient, it reads every review, it negotiates the contract clause by clause, and it coordinates with a million copies of itself for free.
What happens next is the first law doing its work. How fast, and who pays — those took me longer to see.
Law II: Acceleration equals force divided by mass.
Alex Sacerdote, who founded Whale Rock Capital, once commissioned the historian Horace Dediu, a collaborator of Clayton Christensen’s, to chart a century of technology adoption curves. The useful finding is a pair: the radio and the dishwasher.
The radio reached nearly every American household in about seven years, one of the fastest adoptions ever recorded. The dishwasher took decades. Same country, same households, same rising prosperity. The radio you bought, carried home and switched on. The dishwasher had to be plumbed into the back end of the house.
That’s the whole variable. What must be plugged into existing systems adopts slowly. What doesn’t, rips.
Newton’s second law, as every schoolchild meets it: acceleration equals force divided by mass. Hold the force constant and the outcome is decided by the mass of the thing being pushed.
Almost every argument about how fast AI disrupts work is really an argument about F. How capable are the models, how fast are they improving. The radio and the dishwasher say we are staring at the wrong variable. The force is roughly constant across the economy; it’s the same models in the same browser everywhere. What varies is mass: the plumbing, the integration, the procurement cycle, the IT department standing in front of the dishwasher inlet with a clipboard.
Enterprise software was all dishwasher. It still grew thirty to fifty per cent a year, despite having to be wired into incumbent systems and argued past cautious IT departments. AI’s consumer surface is a radio. In Sacerdote’s words, you just open up the browser and it’s there, which is why he says the adoption curve isn’t an S at all but an L stood upright. Vertical first, questions later.
So the second law hands you a tool you can point at any corner of the economy: don’t ask how good the model is, ask how much the corner weighs.
The lightest thing in the economy is an individual knowledge worker, who needs no procurement cycle and no security review, only a browser tab.
And the first measurable acceleration is already in the labour data. For as long as records have been kept, fresh graduates had lower unemployment than older workers. Since 2023 that has flipped, and the gap has kept widening. Shah’s explanation is specific. Entry-level work — structured, supervised, bounded — is exactly what the models already do well, at a fraction of the cost, so firms stop paying the training toll. Why carry a junior at a loss for five years when the model clears that bar today?
He frames it in a way I keep turning over: the entry-level worker is a live proxy for the current frontier of AI capability. Wherever almost-as-good-and-far-cheaper currently bites, that is where the frontier stands. Today it bites at the bottom rung.
The frontier climbs.
Law III: Every action has an equal and opposite reaction.
This is the law everyone forgets, in physics and in economics alike. Nothing is removed, only relocated. Friction, it turns out, is conserved, and when you strip it out of one layer of the economy the right question is where it reappears.
I can see three places.
The first is physics. Actual physics.
For forty years the data centre commoditised. Compute demand grew twenty-five to forty per cent a year, but Moore’s Law delivered that for free, so no component needed real innovation and every part of the machine — boards, memory, networking, enclosures — raced to the bottom. Sacerdote calls what resulted a disaster industry, names left for dead since 1999.
AI broke the bargain. Workloads now grow roughly tenfold a year, which pushes every component in the rack to its physical limit. He calls it the de-commoditisation of hardware, and his case study reads like a parable. Celestica: a contract manufacturer trading at eight times earnings in an industry nobody had thought about for two decades, which turned out to be the sole supplier of Google’s TPU servers and to hold half the market in cloud ethernet switches. The old server was a five-thousand-dollar throwaway. The AI server is a three-hundred-thousand-dollar liquid-cooled machine where a single failure takes down the whole rack. The supplier, in Sacerdote’s phrase, goes from commodity vendor to the maker of a critical part on a plane. You’ll never get swapped out.
The same reversal runs up and down the stack. Circuit boards going from ten layers to forty. Fibre, copper-clad laminate, power supplies — categories that grew five per cent a year for a generation now compounding at thirty-five to fifty, with rising margins and four-year design visibility where the old terms of trade were “we’ll call you next week.”
The two halves of this are one event. The scarce thing in the economy used to be coordinating people, and an entire software industry grew rich taxing it. That scarcity is going to zero. The new scarce thing is coordinating electrons, photons and heat, and the value is flooding back into the physical layer everyone had written off. The friction left the org chart and reappeared in the rack.
The second place is judgment.
Sacerdote’s fund has compounded at roughly forty-four per cent a year, which earns him a hearing on what AI has changed inside it. His answer: less than you’d think. The models get analysts up to speed on ABF substrates in an afternoon and write much better quarterly notes. But, his phrase, there’d better be a really good paragraph on top, which is the wisdom.
His shortest version of it: the AI can be a great reporter; it can’t pick into the future.
What still wins the firm’s internal research award is what won it in Philip Fisher’s day, scuttlebutt. Thousands of face-to-face management meetings a year. The award-winning call on AppLovin came from two analysts who tracked the company while it was still private, learned every competitor and every term of art, flew to the app-advertising conferences in Vegas and Cannes, and built a relationship with the chief executive. “I don’t see AI doing that.”
Something has happened to the price of that paragraph, though. The reporting underneath it used to take a team and a fortnight; it now costs pennies and takes a minute. When the reporting becomes free, the paragraph on top is what gets expensive.
The third place is the one nobody wants to price.
In 1850 Frédéric Bastiat wrote the essay that contains most of economics, What Is Seen and What Is Not Seen. The seen is the shop window; the unseen is everything the money would have done elsewhere. Every fallacy in the subject, he argued, comes from counting the first and forgetting the second.
This was the Curious Mind take on it.
The seen part of this story is good news, and real. Agents strip out the intermediation tax and the surplus flows back to households. Every listing found, every review read, every contract negotiated clause by clause. The toll you have paid your whole adult life gets handed back.
The unseen part: that toll was someone’s income. And in a consumer economy, that someone is also the customer.
Shah’s investment framing is that the AI trade contains an embedded synthetic short on the American consumer, and that the two sides are causally linked rather than merely correlated. The semiconductor complex rises every day because it substitutes for human labour, and America’s economy runs on the spending of its highly paid white-collar workers.
The layoffs themselves are not the dangerous part. The dangerous part is the woman who keeps her job. In March she watches the accountant two desks down get walked out. In May, the ops manager. Nothing happens to her, and that’s the point — nothing has to. That evening she logs into her pension portal and nudges the contribution up two per cent. The kitchen renovation gets postponed. The car can run another two years. Multiply her by forty million and you have a demand shock no layoff statistic will ever show. It is fear, not unemployment, that depresses demand, and it feeds on itself, because softer demand makes firms cut deeper. The displaced spill into blue-collar work and depress that balance too.
Then the stabiliser fails. The central bank’s lever works on cyclical fear: cut rates, money gets cheap, confidence returns, spending resumes. It does not work on her, because what she fears is not this year’s downturn. It is her own obsolescence, and there is no interest rate for that. Shah is blunt about it: all of the traditional heuristics around how economic cycles work will stop applying.
If a productivity boom running alongside a demand depression sounds impossible, it has already happened. The economic historian Alexander Field spent a career assembling the evidence that the 1930s were the most technologically progressive decade in American history. Electrification finally reached the factory floor. The tractor remade the farm, the assembly line remade everything else, and productivity grew faster than in any decade before or since. All of it while a quarter of the workforce stood in breadlines, because the gains arrived faster than they were shared, and frightened households would not spend. The boom and the depression were not opposites. They lived in the same ten years. And the exit, when it finally came, was fiscal policy at the scale of a world war.
The windfall and the wound are the same transaction, seen from its two ends.
What the laws don’t govern
Newton’s laws didn’t make anything move differently. The Principia described what had always been true, visible only once you could imagine friction away. The economy is now running that thought experiment with real money and real lives, and the early readings are in: the flipped youth-unemployment curve, the vertical adoption line, the disaster industry trading like a defence contractor.
Each law fits in a sentence. Every toll persists until something arrives that doesn’t pay it. Speed is set by plumbing, not capability. Nothing is removed, only relocated — out of the org chart and into the rack, out of the intermediary’s revenue and into the household’s nerve.
I should admit what Newton’s admirers learned the hard way. The laws held for two centuries, and then they broke at the extremes — too fast, too massive — and it took Einstein to repair them. These three will have their extremes too, places where the analogy snaps and something stranger takes over. I don’t know where they are. Nobody does yet.
Having lived in the finance world for almost thirty years I think about that the toll-collection booth weekly, after all a lot of the industry is just that.
Bastiat would tell you to watch the unseen column, and he’d be right. But I keep returning to the toll booth itself. For your whole working life you have stood on both sides of it at once, paying tolls with one hand and drawing an income that was someone else’s toll with the other. The frictionless body doesn’t know which side you’re standing on. It just stops paying.
Newton’s gift was subtraction. Take friction away, and what is left is what was always true: bodies in motion stay in motion. The economy is performing the same subtraction on itself — taking away the meetings and the middlemen, the unread contracts, the juniors carried at a loss — and when it is finished, what is left will be what was always true about us.
The machines will do the finding and the summarising — every document read, every comparison run, every report written, tireless and very nearly free. What remains is the last human step: looking at all of it and deciding what it means, which way the future breaks, and whether it should. That step was never machinery. No law of motion governs it. None ever did.
Friction was what we were paid for. It was never what we were for.
A small ask. The Curious Mind has no ads and no algorithm behind it — it grows the way the best ideas always have, by scuttlebutt. If these three laws gave you something to think with, press the heart so strangers can find this, and forward it to one person who'd argue with it well.
Sources: Alap Shah on Hidden Forces with Demetri Kofinas (recorded 25 May 2026); Alex Sacerdote on Invest Like the Best with Patrick O’Shaughnessy (2026); Coase, “The Nature of the Firm” (1937); Bastiat, “What Is Seen and What Is Not Seen” (1850); Alexander Field, “A Great Leap Forward” (2011).
If you are thinking about diving deeper, these other essays are worth visiting: Learning To See Again and The Samurai and The Machine Gun.
These articles are one of the highlights of my working week in finance. Wonderfully insightful and thought provoking. Thank you so much for sharing your thinking in such an accessible way.
Brilliantly researched and put together as ever. I would be fascinated to hear what you think governments of the world plan to do about it, and what the results will be......... thank you