Three rooms.
First room. 138 BC, the fields of Apulia. A smallholder has returned from military service. Seven years away. Olive trees stand in his field — trees that should not be there. He asks a neighbor. The neighbor is no longer a neighbor.
He is a manager for a large estate. The smallholder's land has already been absorbed into a latifundium. While he fought for Rome, Rome swallowed his land.
Second room. 1800, Bolton in Lancashire. A Sunday morning. A handloom weaver walks to church. Twenty-five shillings a week. He works on his own loom, in his own home, at his own pace. Five years of apprenticeship earned him this skill. His wife spins thread beside him. This is prosperity. This is dignity.
He does not yet know that in twenty-five years, doing the same work with the same skill, his weekly wage will be six shillings.
Third room. 2024, a cafe in Seoul. A translator with twelve years of experience opens a message on her phone. It is short. Her main client is switching to AI translation. Rates will drop sixty percent. "If you are interested in transitioning to quality management (post-editing), please contact us." The job: reviewing the software that replaces her.
Between the three scenes lie 2,162 years. The technologies differ. The economies differ. The languages differ.
The structure is the same.
1. The Formula Emerges
In Chapter 1, we began with a single question. Why do some people build fortunes from change while others lose everything? Across fifteen chapters, we crossed three eras. Rome's roads and latifundia. Britain's steam engines and factories. AI's transformers and Big Tech. Now it is time for the larger question. Does a structure that runs through all three eras actually exist?
Four stages repeated.
Technology detonates productivity. Rome built eighty thousand kilometers of roads and aqueducts, creating logistics at imperial scale. The Barbegal water mill complex produced 4.5 tons of flour per day, feeding 28,000 people. During the Industrial Revolution, the spinning machine multiplied productivity by 1,000 times. In the AI era, LLMs directly or indirectly affect 47 to 56 percent of all jobs.
As Acemoglu has noted, however, full direct substitution accounts for only 5 percent of the economy. The direction is agreed upon; the speed is still debated.
The rate of diffusion accelerates. It took 150 years for latifundia to spread across the Italian peninsula. Fifty years for power looms to grow from 2,400 to 250,000 units. Two months for ChatGPT to reach 100 million users. From Rome to AI, the speed of diffusion increased 900-fold.
Capital concentrates in the hands of the few. Rome's Gini coefficient stood at 0.42 to 0.44. The top 1 percent captured 16 percent of total income. In Industrial Revolution Britain, Engels' Pause took hold: labor productivity rose 88.6 percent while real wages actually fell 5.2 percent. A gap of 93.8 percentage points. The fruits of productivity flowed to capital.
In The Poverty of Philosophy (1847), Marx summarized the dynamic: "The hand-mill gives you society with the feudal lord; the steam-mill, society with the industrial capitalist." Brenner amended this: it was the power relations between classes, not technology itself, that determined the direction of concentration. Piketty offered an even simpler formula. r > g — when the return on capital exceeds the rate of economic growth, inequality deepens automatically.
Yet the IMF's Goes observed the opposite trend in 75 percent of countries. Capital concentration is not automatic. Institutions intervene. In the AI era, the top 10 companies in the S&P 500 command 41 percent of total market capitalization — surpassing the 27 percent at the peak of the dot-com bubble in 2000, a historic high.
Social unrest intensifies. In Rome, smallholders became proletarii. In 133 BC, Tiberius Gracchus attempted land redistribution and was murdered. His brother Gaius met the same fate.
During the Industrial Revolution, handloom weavers' weekly wages fell from 25 shillings to 4.5 — an 82 percent decline over thirty years, from 1805 to 1835. In 1811, the Luddites smashed machines; 12,000 troops were deployed to suppress them. In the AI era, 52 percent of Americans say they feel "more concerned than excited" about AI.
Institutions are redesigned. In Rome, the Gracchi reforms failed. The Republic collapsed and the Principate arrived. In Britain, the Factory Act of 1833 was passed — four inspectors overseeing four thousand factories, an imperfect beginning. But a principle was established: the state could intervene in the labor conditions of private enterprise.
In the AI era, the EU AI Act took effect in 2024. Whether it constitutes effective regulation remains an open question.
This is the formula. Technology → Capital Concentration → Social Unrest → Institutional Redesign.
Carlota Perez drew a similar structure in 2002. Technological revolutions pass through an "installation period" and a "deployment period." During installation, financial capital floods in and bubbles form. When the bubble bursts, a turning point arrives. During deployment, institutions absorb the technology and a golden age begins.
Perez's "Frenzy" corresponds to Stage 2 (capital concentration) in this book. Her turning point corresponds to Stage 3 (social unrest). Her deployment period corresponds to Stage 4 (institutional redesign). Among five prior frameworks, Perez shows the highest degree of alignment.
Perez identified the launch of Arkwright's Cromford mill in 1771 as the starting point of the first technological revolution. The timeline matches this book exactly. The shared insight: it is not the technology itself but the system that defines a revolution.
Peter Turchin arrived at the same point from a different angle. Through cliodynamics — mathematical modeling of history — Turchin identified "secular cycles." Every 200 to 300 years, elite overproduction and popular immiseration combine to trigger political instability.
In 2010, Turchin predicted in Nature that the United States would face a political crisis in the 2020s. A decade later, the prediction materialized. Turchin's elite overproduction corresponds to Stage 2 (capital concentration) in this book. Popular immiseration corresponds to Stage 3 (social unrest). The difference: Turchin treats demographics and elite dynamics, not technology, as the primary driver. Keep that distinction in mind.
Ray Dalio's Big Cycle tracks the rise and fall of empires. Education → Innovation → Competitiveness → Output → Trade → Military → Finance → Reserve Currency. If Dalio charted the fate of empires, this book charts the fate of individuals within them. The two are complementary.
2. Time Compresses — But Not Uniformly
Place the three eras side by side and the compression of time becomes visible.
The time it took to recognize the problem. In Rome, 60 years passed before the dangers of latifundia were acknowledged. In the Industrial Revolution, 26 years before the problems of the factory system were recognized. In the AI era, the "GPTs are GPTs" paper by Eloundou et al. appeared just four months after ChatGPT's launch. Recognition accelerated 120-fold.
The time to a first nominal response also compressed. In Rome, 67 years to the Gracchan land law of 133 BC. In the Industrial Revolution, 33 years to the first Factory Act of 1802. In the AI era, one year to the U.S. Executive Order of October 2023. A 67-fold acceleration.
So far, this is encouraging. We see faster. We react faster.
The problem is what comes next.
In the Industrial Revolution, 64 years elapsed before the first effective regulation — the Factory Act of 1833. Rome yields a similar figure. From Crassus's private fire brigade (53 BC) to Augustus's public fire service, the Vigiles (AD 6): 60 years. A striking convergence. Two eras, two entirely different societies, and effective institutional response took 60 to 64 years in both. The gap between them is just three years.
Can this lag be compressed in the AI era? As of 2026, no country can claim to have achieved effective regulation. The EU AI Act has been enacted, but full enforcement is slated for 2026 to 2027. At the U.S. federal level, there is no comprehensive AI regulation.
Recognition accelerated 120-fold. There is no evidence that action has accelerated at the same rate. This is the Recognition-Resolution Asymmetry.
Why does recognition speed up while action does not? Recall the structural causes identified in Chapter 15. Resistance from vested interests is timeless. In Rome, senators were large landowners. They were never going to vote to diminish their own assets.
During the Industrial Revolution, factory owners doubled as Justices of the Peace. The targets of regulation were also its enforcers. In the AI era, Big Tech's annual lobbying expenditure in the United States exceeds $70 million.
Technical complexity compounds the problem. "Do not employ children under nine for more than ten hours" is a relatively simple regulatory design. "Keep algorithmic bias within acceptable limits" is inherently complex.
The EU AI Act's initial draft (2021) was written before ChatGPT existed. By the time the law was adopted in 2024, provisions for foundation models had to be added in haste. While regulation chases technology, technology has already moved on to the next stage.
In the Industrial Revolution, the first effective response across four institutional domains — factory law, trade unions, company law, and electoral law — converges remarkably. The average: 64.2 years. The standard deviation: 8.2 years. Full institutionalization took an additional 105.8 years on average. Institutions do not change quickly. This is the historical base rate.
3. The Dematerialization of Leverage — The Weight of the Tool Decreases
The four-stage cycle repeats across all three eras. But within the cycle, one thing changes systematically. The physical substance of the leverage wielded by the Discerning is disappearing.
Recall Crassus from Chapter 5. Land, construction slaves, silver mines, a fire brigade. The average tangibility of these assets was 74 percent. They had weight. They could be counted. Arkwright, from Chapter 9, was different. A factory system and a licensing network. Tangibility drops to 40 percent. The decisive evidence is the patent revocation of 1785. After the courts stripped Arkwright of his core tangible asset, his business accelerated. Take away the machine, and the system remains untouchable.
At his death in 1792, his estate exceeded £500,000. By his son Richard Arkwright Jr.'s death in 1843, it had grown to approximately £3.25 million — a 6.5-fold increase over roughly fifty years. The system kept generating wealth long after its creator was gone.
The AI native builds a business with questions. Midjourney has 130 to 163 employees. Its marketing budget is close to zero. Its office is Discord. Its servers are rented cloud infrastructure. Average tangibility drops to 12 percent. The source of wealth is the cognitive combination itself — the question of which problem to solve, and how.
74 percent → 40 percent → 12 percent. The dematerialization of leverage. From what can be counted to what can be described. From what can be described to what is difficult to describe.
The Arkwright Test applies this logic directly. Strip each era's Discerning of their core tangible asset. What happens?
Crassus. During the proscriptions, other senators had their assets seized. Crassus was the one buying. In the era of tangible leverage, "being stripped" and "stripping others" operate through the same mechanism. Result: partial pass.
Arkwright. After the 1785 patent revocation, his business accelerated. Making the technology public expanded the market, and Arkwright's system could not be replicated. Result: full pass.
The AI native. In August 2022, Stable Diffusion was released as open source. The same functionality as Midjourney became available for free. Midjourney's revenue grew roughly tenfold, from $50 million in 2022 to $500 million in 2025. Open the code, and the cognitive combination remains irreplicable. Result: pass.
As the weight of the tool decreases, the weight of the outcome increases.
4. The Democratization of Entry Barriers — The Gate Opens, and a New Wall Appears Inside
The minimum cost of entry to becoming one of the Discerning has collapsed dramatically.
For Crassus to organize a private fire brigade and buy up proscription properties, he needed initial capital of 300 talents — about 8,000 times the annual pay of a soldier (900 sesterces). The number of people with access to that kind of capital was limited to 300 senators and a few thousand equites.
Arkwright started the Cromford mill for about £500. The partnership with Jedediah Strutt was decisive. A skilled artisan's annual income was roughly £50, so the entry cost was about 10 times median earnings. You did not need to be a senator. A wig-maker could get in.
The minimum entry cost for an AI native is a $20 to $200 monthly API subscription. About $240 per year. That is 0.003 times the U.S. median annual income of $60,000. A laptop and an internet connection are, in theory, all you need.
8,000x → 10x → 0.003x. A total decline of 2.7 million times. This figure should not be read as a precise multiplier. It conveys scale. The currency units, purchasing power, and economic structures are entirely different.
Stop here and it is a beautiful story. Entry barriers vanish, and anyone can become one of the Discerning. Reality differs.
After Arkwright's patent was revoked, anyone could build a water frame. Scottish immigrants in Massachusetts even secured U.S. government support to reproduce Arkwright's machines. The result was "coarse and irregular fabric." Accessing the technology and building the system are different problems.
The same holds in the AI era. ChatGPT and Claude cost $20 a month to access. The failure rate for AI startups is 90 percent. According to MIT research, 95 percent of AI pilot projects fail. A structural gap exists between tool access and leverage acquisition.
When the gate opens, a new wall appears inside. Tangible barriers like capital and social status (Rome) gave way to intangible barriers like organizational capability (the Industrial Revolution). Now hyper-intangible barriers — cognitive adaptability — have emerged (the AI era).
And new kinds of barriers are added. The attention barrier: when every tool is free, "what to focus on" becomes the barrier. The execution speed barrier: in a world where Cursor reached $100 million in annual revenue within twelve months, speed itself is competitive advantage. The trust barrier: in a world where anyone can build, the things that can be trusted are few.
The paradox of democratization. Even when entry costs fall 2.7 million times, the probability of success does not rise proportionally. The nature of the barrier changes. The barrier itself does not disappear.
5. The Evolution of Moral Complexity — The Face of the Perpetrator Blurs
As the profile of the Discerning changes, the moral landscape shifts with it.
As we saw in Chapter 5, a modern reader encountering Crassus's fire-sale extortion has no difficulty reaching a moral judgment. The face of the perpetrator is clear. Arkwright, from Chapter 9, is more complicated. Child exploiter and job creator at once. The question was contested even in his own time.
The Discerning of the AI era exist inside an API. Shopify's CEO sent a company-wide memo in April 2025: "You must demonstrate that a task cannot be done by AI before requesting a new hire." Fiverr's CEO declared "AI is coming for your jobs" while cutting 250 positions.
Who took the translator's job? The engineer who built the API? The CEO who adopted it? Everyone who clicked the $20 subscribe button?
The perpetrator has been distributed. Crassus's fire-sale extortion is intuitively immoral. Arkwright's child labor was contested even in his day. When an AI-native founder's code editor automates a legal assistant's work, the question "who is immoral?" resists an easy answer.
This distribution makes institutional response harder. The more precisely a target can be identified, the easier regulation becomes. In Rome, it was possible to identify large landowners. The Licinian law set a ceiling of 500 iugera. During the Industrial Revolution, it was possible to identify factory owners. The Factory Acts imposed obligations on them.
In the AI era, "whom to regulate" is itself the debate. The EU AI Act sidestepped the problem by classifying AI systems by risk level — categorizing the technology, not the people. Whether this approach works remains to be seen.
A villain who happened to be a genius (Crassus). A disruptive innovator (Arkwright). An unintentional destroyer (the AI native). Direct exploitation decreases. Moral ambiguity increases. The face of the perpetrator grows steadily more blurred. This is the moral predicament of our era.
Let us bind the three dimensions together. As leverage becomes intangible, entry barriers fall and the outline of the perpetrator fades. The three shifts point in a single direction. The lighter the weapon of the Discerning, the harder the ripples are to trace.
Crassus's fire brigade was visible. Arkwright's system could be drawn on a blueprint. The AI native's cognitive combination is difficult even to describe. As the weapon grows invisible, the attribution of harm grows uncertain, and institutional response grows intractable.
6. The Formula of the Displaced — What Does Not Change
While the profile of the Discerning shifts from era to era, something in the profile of the Displaced stays constant.
Identity locked to a single skill. Rome's smallholder tied his identity to the land his grandfather had cleared. His civic status as an assiduus — a propertied citizen — was itself contingent on land ownership. Lancashire's handloom weaver internalized his craft through five to seven years of apprenticeship. As E.P. Thompson recorded, he was an artisan who "placed a value on working without a master."
The AI-era translator built an identity over ten to fifteen years of professional experience as "a bridge between two languages."
Position in the execution layer. The smallholder worked the land himself. The weaver sat at the loom himself. The translator translated herself. All three occupied the execution layer of the value chain. None occupied the design layer.
Underestimation of the speed of structural change. The smallholder could not track changes in his hometown during six or seven years of military service. In 1808, fifteen thousand weavers petitioned Parliament — and were dismissed. They had underestimated the speed at which power looms would spread.
In the AI era, 61 percent of workers say they are "considering" upskilling, but only 4 percent are actually pursuing it. A 57-percentage-point gap between intention and action.
And then the cruelest constant. The irony of replacement. Rome's smallholders expanded the empire through military service. Wars of conquest produced prisoners of war. Prisoners of war became slaves on the latifundia. The latifundia displaced the smallholders. The smallholders built their own replacement with their own hands.
The same structure applies to the handloom weaver. Arkwright's spinning innovation supplied cheap thread. Cheap thread created the weavers' golden age (1790 to 1810). The same industry's technology gave them prosperity, then the power loom destroyed it.
The AI-era translator performs post-editing — correcting errors in AI translation. That correction data makes the AI more accurate. "Training the software that will replace me." The structure is the same. The Displaced accelerate their own displacement.
Only the speed of decline accelerates. 150 years (Rome, two to three generations) → 30 years (the Industrial Revolution, one generation) → 3 years (the AI era, ongoing). A 50-fold acceleration. The structure does not change. Only the clock speeds up.
Why four stages, specifically? Is anything missing?
The most plausible fifth candidate is international competition. Rome's expansion was driven by its rivalry with Carthage. During the Industrial Revolution, Anglo-French competition accelerated technology investment. In the AI era, U.S.-China competition shapes the speed and direction of regulation.
But international competition is an environmental variable, not a stage within the cycle. It does not occur sequentially inside the cycle. It is an external condition that accelerates or deforms all four stages.
Cultural resistance is another candidate. Smallholders invoked ancestral custom. The Luddites smashed machines. In the AI era, discourse about "uniquely human" capabilities is spreading. But cultural resistance is inseparable from Stage 3 (social unrest). Economic protest and cultural backlash have always appeared intertwined.
The number could have been three or five rather than four. It is four because four is the minimum model derived from observation. Collapse the sequence to two stages — technology → institutions — and the intermediate mechanisms of capital concentration and social unrest disappear. Add more stages and the explanatory power weakens further at n = 3. This formula is not optimal but minimal. A model that is minimally sufficient is optimal.
7. A Warning Against Excessive Neatness — What This Formula Is Not
To stop here would be dangerous.
The four-stage cycle repeats neatly across three eras. Too neatly. Neatness itself is a warning sign.
Karl Popper cautioned in 1957, in The Poverty of Historicism: there are no laws in history. There are only trends.
History is a process that occurs only once. Deriving universal laws from a singular process is impossible. Human action cannot be predicted with certainty. Since future knowledge determines future history, prediction of the future is logically impossible as well.
Nassim Taleb extended this critique into cognitive science. The narrative fallacy — "our limited ability to look at sequences of facts without weaving an explanation into them." A story that fits is not necessarily a story that is true.
A black swan — an extreme event outside the range of prediction — can neutralize any pattern analysis.
We must acknowledge the possibility that this book's four-stage cycle is a pattern imposed after the fact.
Three cases cannot establish a universal law. n = 3. This is not a sample size that permits claims of statistical significance.
Counterexamples exist. Not every technological change produces the four-stage pattern. The Neolithic Revolution was a gradual transition spanning millennia. There were no institutions to redesign. The printing press (1440s) triggered enormous social upheaval — the Reformation, the Scientific Revolution — but printers were small-scale operators and capital concentration was limited.
During the Green Revolution (1960s-70s), international organizations proactively managed technology diffusion, and the cycle was dampened. Electricity (1880s-1920s) complemented labor rather than replacing it. It created more jobs.
The counterexamples reveal the cycle's operating conditions. When technology replaces labor. When capital concentration is intrinsic to the technology. When institutions fail to respond proactively. When all three conditions are met simultaneously, the cycle operates. When they are not, it does not.
Counterexamples are not the only variant. There are also compressed cases.
South Korea's Miracle on the Han River is an instance where the four-stage cycle was compressed into 35 years. The technology was not indigenous but transplanted. POSCO, Hyundai's shipyards, Samsung's semiconductors. Arkwright's Cromford, as if designed by a government planning office. Capital concentrated in the form of the chaebol. Social unrest erupted in compressed waves, from the labor movements of the 1970s to the June Democratic Struggle of 1987. Institutional redesign proceeded rapidly after democratization.
Compression is possible. But compression carries a cost. The faster the cycle runs, the denser the shock. If recognition has accelerated 120-fold in the AI era, institutions may also accelerate — as South Korea compressed the cycle to 35 years. But who pays the cost of that compression remains an open question.
The direction of causation is not straightforward either. This book presents a sequence of technology → capital concentration. The reverse direction is also possible. Rome's military conquests (political power) enabled infrastructure investment. Roads were built for legions.
During the Industrial Revolution, the accumulation of colonial capital enabled factory investment. In the AI era, Big Tech's preexisting monopolistic positions in the 2010s enabled AI research and development spending.
Joel Mokyr argued that technology determines institutions. Daron Acemoglu and James Robinson argued that institutions determine technology. In practice, the relationship is bidirectional reinforcement. Technology is not a cause. It is a catalyst. It opens possibilities but does not determine direction.
The uniqueness of each era also demands respect. Roman slavery, the Industrial Revolution's energy transition, and AI's cognitive displacement cannot be reduced to this framework.
The Roman economy was slave-based. Smallholders were replaced not by wage workers but by slaves. The Industrial Revolution was a transition from biological energy to fossil fuels — a singular event in human history. AI targets cognitive labor for the first time. A qualitatively different transition.
This formula is not a law. It is a lens.
A lens sharpens certain features at the cost of blurring others. Chance, individual choice, black swans, institutional path dependence — acknowledging what this lens cannot show makes it more useful, not less.
8. Why the Pattern Is Still Useful
A lens is not a law. But seeing without a lens is more dangerous than knowing a lens has limits.
Popper rejected the deterministic prediction of history — the claim to know "what will happen next." This formula makes no such prediction.
Extracting questions from historical patterns belongs to a different epistemological category than predicting the future. Just as a physician reads a patient's medical history for warning signs without claiming to prophesy the future, drawing "questions we should be asking right now" from historical patterns is diagnosis, not prophecy.
What this formula does is force questions.
If technology is detonating productivity, where is capital concentrating? The top 10 companies in the S&P 500 hold 41 percent of total market capitalization. The three major cloud providers (AWS, Azure, GCP) command 63 percent of the infrastructure market.
The training cost of GPT-4 ran $78 million in compute alone, and is projected to surpass $1 billion by 2027. Usage is being democratized, but development is being oligarchized. Forcing that question is the formula's value.
Where is social unrest appearing? According to Acemoglu and Restrepo, automation explains 50 to 70 percent of U.S. wage inequality. Each additional robot reduces employment by six workers and depresses wages by 0.42 percent. These figures apply to physical automation. The impact of cognitive automation has not yet been measured at scale.
Are institutions keeping pace? The historical base rate is 60 to 64 years. Will that base rate compress in the AI era? Accelerating factors exist: real-time information, existing educational infrastructure, global benchmarking, the use of AI itself as a regulatory tool.
Decelerating factors are equally powerful: regulatory capture, technical complexity, global competition, political polarization. Recognition has accelerated 120-fold, but there is no guarantee that action will accelerate at the same rate.
These questions are equally valid for Rome's Gracchi, for the Industrial Revolution's Arkwright, and for us in 2026.
Acemoglu and Robinson pointed to the case of Nogales — two cities of the same name straddling the U.S.-Mexico border. Same geography, same climate, same cultural roots. Per capita GDP on the American side is roughly three times the Mexican side. The difference was institutional. The same technology produces different outcomes under different institutions.
The 2024 Nobel Prize in Economics was awarded for that insight.
The same experiment is underway in the AI era. The EU has chosen rights-based regulation. The United States has chosen innovation first. China has chosen state control. One technology, three different institutional frameworks. Who is right remains unknown.
Technology opened possibilities; capital and institutions determined the direction. That sentence is this book's core argument. It is not technological determinism. Technology is a catalyst. Direction is ours to decide. The fact that the decision has not yet been made is the defining feature of this era.
9. The Asymmetry of Acceleration — What the Formula Reveals
Let us compress the formula into a single table.
Some things accelerate. Technology diffusion — from 150 years to roughly two months, 900-fold. Capital accumulation — from a CAGR of 11 percent to 37 percent; AI startups grow by hundreds of percent annually. The speed of decline — from 0.7 percent per year to 30 percent, a 5- to 8-fold increase each era. The speed of problem recognition — from 60 years to under one year, 120-fold.
Some things do not accelerate. Effective institutional response — 60 years in Rome, 64 in the Industrial Revolution, unresolved in the AI era. Individual psychological adaptation — identity lock-in follows the same pattern across all three eras. Moral consensus formation — with the distribution of perpetrators, this is actually becoming harder.
This is the Asymmetry of Acceleration. Technology and capital accelerate with each era, but institutions and individual adaptation do not accelerate at the same rate. In that gap, the Discerning build fortunes and the Displaced pay the price.
Revenue per employee among AI natives quantifies the gap. The S&P 500 average is $550,000. The AI-native average is $3.48 million — 6.3 times higher.
Cursor achieved $1 billion in annual revenue with 300 employees. $3.3 million per person. A world where 10 people do the work of 100 has already arrived. For the other 90, a different world has arrived.
10. One Person Across Three Eras
Spring 2026. A study in Seoul.
An investor is reading this chapter. The formula is not an academic curiosity but a practical question. Where are we now in the four-stage cycle?
Stage 1 (Technology → Productivity Explosion) is clearly underway. LLMs affect 47 to 56 percent of all jobs. Direct substitution, however, accounts for only 5 percent.
Stage 2 (Capital Concentration) is also underway. The Magnificent Seven drive 66 percent of S&P 500 earnings growth.
Stage 3 (Social Unrest) is showing early signs. Fifty-two percent of Americans express concern about AI. Translators' incomes have fallen 60 to 80 percent. The gap between the intention to upskill (61 percent) and actual pursuit (4 percent) is widening.
Stage 4 (Institutional Redesign) has not yet arrived.
According to the historical base rate, effective institutional response takes 60 to 64 years. There is hope that this lag can be compressed in the AI era. At the same time, there is the reality that recognition has accelerated 120-fold without action accelerating at the same rate.
The investor asks: Am I among the Displaced, or the Discerning?
The formula of the Displaced has not changed. Pride in a single skill. Complacency in the execution layer. Underestimation of the speed of change. When all three conditions are met simultaneously, regardless of era, an individual becomes a casualty of structural transformation.
The formula of the Discerning is the inverse. A portfolio of multiple capabilities. A position in the design layer. Early recognition of structural shifts.
But one thing must be added. Survivorship bias. Crassus's success depended on the structural luck of his political connection to Sulla. Arkwright's success depended on the luck of a partner named Strutt and the waterpower at Cromford.
A man named Thomas Highs developed the same technology at the same time as Arkwright. Lacking capital and business acumen, he was forgotten by history. The five-year survival rate for new businesses in the United States is only 51.6 percent. Seventy to eighty percent of venture-backed startups fail to reach their target returns.
Reading the change is a necessary condition, not a sufficient one. Those who fail to read rarely succeed. Those who read do not necessarily succeed either. This asymmetry is the formula's practical lesson.
Transition — From Formula to Framework
A structure runs through three eras. Technology detonates productivity. Capital concentrates. The Displaced appear. Institutions follow.
This structure is not a law but a lens. Three observations cannot establish a universal law. Counterexamples exist, the direction of causation is complex, and each era's uniqueness exceeds the framework's reach.
Yet the lens is useful. A map is not the territory. Knowing the map's limits is better than traveling without one.
We have used this lens to look at the past. Now it is time to look at the future.
In the next chapter, the formula becomes an investor's framework. Where do we stand in the cycle? How should we position for the next twenty years? Distinguishing what history can tell us from what it cannot.
End of Chapter 16. Next: Chapter 17 — The Investor's Framework: Positioning for the Next 20 Years, as Told by History