Preamble
On the sensation of standing at a hinge of history
There is a sensation familiar to anyone who has spent serious time studying history and then finds themselves living through a moment that resembles, in its structural anatomy, something they have read about but never expected to witness. It is not quite deja vu. It is more like the recognition of a musical key — you have heard the melody before in different instrumentation, in a different room, performed by different hands, but the underlying harmonic relationship is unmistakeable. The chord sequence is the same. Something large is turning.
This essay is an attempt to map what is turning, and why, and for whom it is visible. It works across several registers simultaneously — political, monetary, technological, psychological, and generational — because the phenomenon under analysis does not respect disciplinary boundaries. The institutions failing, the technology accelerating, the money corroding, the crowds responding, and the people who can and cannot see it: these are not separate stories. They are one story told from different angles, and the full picture only becomes legible when all the angles are held at once.
The inflationary system that survives by leaking value, managed by institutions that survive by controlling information, is now colliding with a deflationary technology that cannot leak and a distributed ledger that cannot be controlled.
We are living through the intersection of two exponential curves moving in opposite directions. Institutional legitimacy has been declining exponentially, its failures compounding, its self-protective behaviours becoming more visible and less justifiable. Simultaneously, technological capability has been rising exponentially, destroying the information asymmetries on which institutional power has always depended. Where these two curves cross is exactly where we are standing in February 2026. That crossing is not paranoia. It is geometry.
The information asymmetry that has underwritten institutional power for all of recorded history is ending. Not gradually. Exponentially. The institutions that depend on it are responding with increasing desperation and decreasing competence.
A critical variable, too long absent from this analysis, is the human one: the crowds moving through this transition are not rational actors calmly updating their priors. They are narrative-driven, status-sensitive, mimetic creatures whose behaviour under stress diverges sharply from calm theoretical models. The revised edition incorporates this dimension fully — and it changes the picture in ways that matter for how any individual should think about preparation.
Chapters · Select to expand · Select a chapter to expand
Here is a pattern. See if it sounds familiar.
A government institution is created to solve a problem — housing, healthcare, law enforcement, financial regulation. Over time it grows. It adds layers of management, compliance, oversight, reporting. Each layer was added for a good reason. Individually, none of them are unreasonable.
But collectively, they produce something strange: an organisation that now spends most of its energy managing itself rather than solving the original problem.
That is the NHS spending more per patient every year while waiting lists grow. That is a planning system so complex that Britain builds fewer homes than almost any comparable country — not because we lack land, money, or people who want homes, but because the permission system has become a machine for producing its own paperwork.
A historian called Joseph Tainter studied how civilisations collapse and found a consistent pattern. They do not usually fail because of invaders or natural disasters. They fail because complexity becomes its own reward. The institution exists to protect the institution. The problem it was created to solve becomes secondary.
We are seeing that now. Everywhere. At the same time.
But here is the part that is new.
When institutions reach this stage historically, they possessed one major advantage: they could control the story. They could delay, deny, manage the narrative, keep the embarrassing documents in filing cabinets where nobody would find them.
That advantage is gone.
The Epstein document releases are a precise example. Nobody found a secret. The information was in government systems for years. What changed is that the tools to surface, share and analyse it became more powerful than the tools to suppress it. The grooming gangs court archive deletion is the same pattern — an institution using administrative tools to protect itself from accountability for its own failures. Except now everyone can see them doing it, in real time.
The behaviour has not changed. The visibility has. And visibility changes everything.
The Gramsci Dimension
Antonio Gramsci, writing from a fascist prison in the 1930s, described what he called an organic crisis — a moment when the ruling order loses not just competence but legitimacy. Not the ability to govern by force, but the consent that makes force unnecessary.
His interregnum is the interval between: the old world still structurally present but no longer believed in; the new world not yet coherent enough to replace it. "The old world is dying, the new world struggles to be born: now is the time of monsters."
That is not a metaphor for the current moment. It is a description of it.
The trust data confirms this analytically. Edelman and Gallup have tracked institutional confidence across Western democracies for decades. The trajectory is not ambiguous: governments, media, central banks, courts, global institutions — all below 50% trust in most measured populations, all on continuous downward trends since approximately 2008, none showing reversal.
This is not a political problem amenable to electoral solution. It is a structural condition — the Tainter mechanism (complexity without return) operating through the Gramscian channel (legitimacy withdrawal). The institutions are failing and the populations know it and there is no mechanism within the existing system for correction.
The Information Order
The additional layer — the one that makes this moment different from previous periods of institutional stress — is the collapse of the information asymmetry that historically sustained institutional authority.
Institutions have always made errors. What changed after roughly 2008, and accelerated sharply after 2016, is the capacity to surface those errors at scale, in real time, with verification. The tools of transparency have outrun the tools of control.
Orwell understood that information control was the primary instrument of power — not force, but the management of what is thinkable. The current moment is one in which that management is failing structurally, not through any single act of resistance, but through an architectural shift in the information environment.
You cannot unprint the printing press. The equivalent statement for the present moment is: you cannot re-establish information scarcity in an environment of structural information abundance.
The institutional order that was built on information scarcity is encountering an environment of structural information abundance. That is the thesis of this chapter, stated plainly. Everything else in the analysis of Chapter One follows from it.
Gramsci wrote from a fascist prison that 'the old world is dying, and the new world struggles to be born: now is the time of monsters.' The interregnum — between the death of one order and the establishment of another — is characterised not by orderly power transfer but by dangerous fluidity.
Kindleberger's hegemonic stability theory argues stable international orders require a single dominant power providing global public goods: reserve currency, security guarantees, free trade enforcement. When the hegemon weakens, no one provides the public goods — producing cascading instability.
Strauss and Howe's Fourth Turning model (1997) predicted a Crisis era beginning approximately 2008 and resolving around 2030 — a period of institutional demolition and reconstruction. We appear approximately two-thirds through this arc.
We are in Gramsci's interregnum. The American-led post-war order — dollar liquidity, NATO security guarantees, WTO trade rules, a notional commitment to liberal international norms — is not being replaced. It is dissolving. The distinction matters enormously. Replacement is a transition. Dissolution is an interregnum. And the interregnum is the time of monsters.
China is the most obvious candidate for hegemonic succession, but the timing is structurally premature. The property crisis that began with Evergrande has been administered into prolonged slow deflation — historically the worst available outcome for an economy that depended on construction for approximately a third of GDP. Demographics are darkening on a 30-year horizon already inside the strategic planning window. Xi's centralisation of authority has the efficiency advantages of decisive command and the fragility disadvantages of single-point failure.
The genuinely novel development — the one that most disrupts the bipolar Cold War analytical framework — is the confident multi-alignment of the middle tier. India, Turkey, the Gulf states, Brazil, Indonesia: these states are not choosing between American and Chinese orders. They are building relationships across previously incompatible alignments, extracting value from their position in the gap between competing great powers, and refusing the loyalty demands of both. This is not Cold War non-alignment from weakness. It is active multi-alignment from growing structural confidence — the recognition that the Western order no longer commands sufficient moral authority or material dominance to demand exclusive loyalty.
Kurzweil's Law of Accelerating Returns posits that the rate of technological progress is itself exponential — each generation of technology produces the next generation faster. Applied to AI, this generates the singularity thesis: the point at which AI capability surpasses human cognitive performance across all domains, enabling recursive self-improvement at speeds exceeding human comprehension.
The critical distinction is between narrow capability gains (AI better at specific tasks) and general capability gains (AI better at the meta-task of getting better). The latter is the phase transition. Evidence suggests we are approaching this threshold.
The rice-and-chessboard illustration: by square 32 the quantities are large but conceivable. By square 64 they exceed total global agricultural output across all of recorded history. The experience of the first half gives you no intuitive preparation for the second. We are past square 32 in AI development.
A clear-eyed assessment of AI in February 2026 requires neither techno-utopianism nor the scepticism that has serially underestimated capability development. The honest position: we are not at the singularity. We are on a trajectory whose mathematical form leads there, and we have crossed the inflection point at which exponential growth becomes visible to careful observers while remaining non-obvious to those applying linear extrapolation to an exponential phenomenon.
Consider the capability delta across 36 months. Models available in early 2023 produced fluent text and competent code. Models available in early 2026 conduct legal research at a level competitive with junior associates; diagnose pathologies in medical imaging at accuracy levels exceeding specialist human performance on specific datasets; generate, debug, and refactor complex multi-language software; conduct original scientific literature synthesis; and engage in multi-step strategic reasoning across extended temporal horizons. This is not linear improvement. The architecture of improvement has changed shape.
More significant than any individual capability gain is the structural recursion now operating. AI systems are being used to generate synthetic training data for the next generation. They are being used to design improved architectures and identify their own failure modes. The thing that is improving is beginning to improve the process of improvement. This is the engine of the singularity thesis, and it has started.
The most rigorous empirical confirmation of this trajectory comes from METR — the Model Evaluation and Threat Research organisation — whose longitudinal study of AI agent capability from 2019 to 2026 provides something rare in this field: measured, reproducible data rather than benchmark scores or anecdotal capability claims. Their methodology is straightforward and therefore powerful. They took the most capable AI agents available at each point from 2019 to 2026 and tested them on approximately 230 tasks — primarily coding tasks with components requiring general reasoning — and measured each task's length in terms of how long human professionals take to complete it. They then measured the 'time horizon': the task length at which the AI agent succeeded fifty percent of the time.
Two patterns emerged with statistical clarity. Task length is highly correlated with agent success rate — R-squared of 0.83 — confirming that the metric is measuring something real and consistent. And the time horizon itself has been growing exponentially, doubling every seven months across the full 2019-2026 period. When ChatGPT launched in late 2022, the frontier agent time horizon was approximately thirty seconds. By early 2026, frontier agents can autonomously complete coding tasks that take human professionals over fourteen hours. That is not a marginal improvement. That is the chessboard moving from square fifteen to square twenty-two.
The Acceleration Within the Acceleration
The finding that should command the most attention is not the doubling rate itself but what happened to the doubling rate. Across the full 2019-2025 period, time horizons doubled every seven months. In 2024-2025 alone, the doubling rate compressed to every four months. The exponential is itself accelerating. This is the empirical signature of the recursive improvement dynamic described in the singularity thesis: when the technology improving is also being used to improve the technology, the rate of improvement accelerates. We are watching this happen in measured, published data.
The extrapolation from the baseline seven-month doubling is already striking: 2027 brings autonomous completion of full work-day tasks, 2028 of full work-week tasks, 2029 of full work-month tasks. If the accelerated four-month doubling from 2024-2025 persists, those dates compress further — month-long autonomous task completion potentially arriving in 2027. The researchers themselves note that the rate might slow, but they also note it might speed up further, and that a superexponential trajectory is consistent with both the data and the theoretical mechanism of AI-assisted AI development.
The implications for labour markets, for professional identity, for the institutional structures built around the assumption of irreplaceable human cognitive work, are not distant or theoretical. A system that can complete work-week-length tasks autonomously does not merely assist professionals. It substitutes for them across a very wide range of cognitive work. The transition from assistant to substitute does not happen gradually at the individual task level — it happens suddenly at the level of the role, when the time horizon of autonomous competence crosses the median complexity of the work the role requires. Many roles will cross that threshold within the planning horizon of people currently in education.
AlphaFold and the Adjacent Science
AlphaFold's solution of the protein-folding problem deserves to be ranked alongside the discovery of the double helix as a moment that changed the shape of biological possibility. Every protein whose structure can now be predicted is a potential drug target that was previously opaque. Drug development timelines are compressing. CRISPR has moved from laboratory curiosity to clinical application for conditions previously considered irreversible. The biology of senescence is yielding to new analytical tools. We are at the beginning of understanding ageing at a mechanistic level — and therefore of understanding how that process might be interrupted.
The most important point about the technological singularity is psychological and mathematical, not technical. Human cognition evolved for linear extrapolation. Our intuitive extrapolation machinery is calibrated for linear trends. This is a profound and systematic error applied to exponential phenomena. The people who seem unbothered are not better calibrated. They are applying the correct algorithm for a linear phenomenon to an exponential one. The dissonance between what careful observers see approaching and what the general discourse acknowledges is not a sign of error. It is a mathematical artefact of position on the curve.
Mises's regression theorem: money must trace its value back to a pre-monetary commodity use. Nick Szabo's 'unforgeable costliness': for sound money, the cost must be physically embedded and unfakeable — incapable of being conjured by decree.
The energy theory of money: money is the technology for storing and transporting human productive energy across time and space. Quality is measured by how little energy leaks during storage and transport. Austrian capital theory holds that sound money is a prerequisite for rational economic calculation — corrupted price signals produce malinvestment on a structural scale.
Schumpeter's creative destruction: new technological paradigms do not add to existing economic architecture — they destroy it, creating crisis during transition and greater productive capacity in the new equilibrium.
What Fiat Money Actually Does
To understand what Bitcoin is, you must first understand with precision what the fiat system does — not as polemic but as mechanism. The post-war monetary order, and in its more radical form the post-1971 order, is a system in which money is created by sovereign authority, managed by central banks whose independence from elected government is nominal rather than structural, and backed by nothing other than institutional credibility and the coercive power of legal tender laws.
The physical constraint on money creation — the gold link that imposed, however imperfectly, some relationship between monetary expansion and productive capacity — was removed definitively in 1971 when Nixon closed the gold window. What replaced it was an institutional constraint: the discipline and credibility of central banks. That institutional constraint has proven, across 55 years of evidence, to be elastic in ways that physical constraints are not. The monetary base of the dollar has expanded by approximately 10,000 percent since 1971. Debt-to-GDP ratios across the developed world have reached levels that in any previous era would be classified as systemic insolvency.
The Cantillon effect is the mechanism by which this matters for the distribution of economic power. When money is created, it appears first at the point of creation: central banks, primary dealers, large financial institutions, governments. From there it diffuses outward through the economy. By the time it reaches wages and consumer prices, early recipients have already deployed it into assets whose prices have already risen. The result is a systematic, structural transfer of real wealth from those who hold monetary savings — predominantly the middle class — to those who hold real assets. This is not a side effect of fiat money. It is a structural feature of the mechanism itself.
The Thermodynamic Proposition
Against this backdrop, Bitcoin makes a specific and radical proposition. Not: here is a better payment system. Not: here is a superior store of value. The proposition is deeper: what if we built money whose supply was determined by mathematics rather than institutional decision, whose ledger was maintained by physics rather than trust, and whose properties were enforceable by anyone with a computer rather than by any authority?
The Proof of Work mechanism is not merely a technical solution to the double-spend problem. It is the encoding of physical reality into a digital monetary system. To add a block to the Bitcoin blockchain, you must perform real computation. Real computation consumes real energy. Real energy corresponds to real productive work performed somewhere in the physical world. The hash that results is unforgeable — not by institutional decree but by the fundamental constraints of mathematics and thermodynamics. You cannot produce a valid Bitcoin block without actually expending the energy. The work is embedded in the money. This is Szabo's unforgeable costliness implemented in digital form.
Bitcoin as Temporal Bridge — Transporting Energy Across Paradigms
The most powerful framing is the temporal one. Money is a technology for moving value through time. When you earn something today and spend it in ten years, you are using money as a time machine — transporting the productive energy you expended now to a point of future consumption. The question for any monetary system is: how much energy is lost in transit?
Fiat money leaks badly and by design. At a modest 3 percent annual inflation — below the actual long-run average in most fiat economies — a stored monetary value loses half its purchasing power in approximately 23 years. At the inflation rates experienced between 2021 and 2024, the leakage was acute and rapid. The person who saved diligently in the currency of their state was paying a continuous and involuntary tax on every unit of productive work they chose to defer. The incentive structure this creates — spend now, do not save, borrow to acquire assets — is precisely the incentive structure of a highly leveraged, consumption-driven, financialised economy. The monetary system produces the economy it incentivises.
Bitcoin is designed to be a lossless temporal container. The fixed supply at 21 million units is not a policy. It is not an institutional commitment. It is a mathematical property of the protocol, enforced by the economic self-interest of every participant in the network. Changing it would require convincing a distributed global network of economically-incentivised validators to accept a change that would directly diminish the value of their holdings. The architecture makes the constraint structurally near-impregnable. The saver is rewarded. The incentive structure inverts: save, produce, defer consumption, accumulate. This is the incentive structure of a low-time-preference civilisation — one that plants trees whose shade it will not sit under, builds infrastructure for grandchildren, invests in long-duration projects.
The Inflationary System Meets the Deflationary Machine
The collision you sense at the macroeconomic level is most precisely understood as the collision of two incompatible technological paradigms operating at different points on their adoption curves. The fiat system is a mature technology past its peak. It works, in the limited sense that any engineering system works when its structural assumptions are met. But it was designed for a world of information scarcity — in which the institutions managing money could rely on information asymmetry between themselves and the public to maintain the credibility that backs the system.
That information environment has been destroyed by the same technological acceleration producing everything else discussed in this essay. When central banks claimed inflation was 'transitory' in 2021, the contrary evidence was immediately, publicly, and inescapably available. The informational scaffolding that sustains institutional credibility is being systematically dismantled by the technology of transparency.
Bitcoin is native to the new information environment. It does not depend on institutional credibility because it has no institutions. All its rules are public, all its transactions transparent, all its supply mathematics verifiable by anyone with a laptop. It arrived, through the mechanism of Satoshi's specific history and the 2008 financial crisis, at precisely the moment when the world was beginning to become a world in which information asymmetry was structurally unavailable. That timing may not be coincidence. It may be the free market — that emergent collective intelligence of freely acting individuals — identifying and filling a gap that technology was simultaneously opening.
The free market, in its deepest sense, is not a political ideology. It is a description of what happens when individuals are free to coordinate without central direction — the emergence of order from distributed decision-making, the aggregation of dispersed knowledge into price signals. What the free market has done, operating through millions of independent decisions, is identify the gap opened by the informational destruction of fiat credibility and fill it with a technology that does not require the credibility that is being destroyed. This is Schumpeter's creative destruction operating at the level of monetary infrastructure. Bitcoin is the transitional vessel — carrying stored energy and productive work across the boundary between the old paradigm and whatever comes next.
Kahneman's dual-process theory distinguishes System 1 — fast, associative, pattern-matching — from System 2 — slow, deliberate, analytical. Expert intuition is the successful training of System 1 on high-quality feedback environments. The danger: System 1 trained on one environment will fire confidently and wrongly when the environment changes.
Taleb's antifragility describes entities that gain from disorder — strengthened, not merely preserved, by exposure to volatility and stress. A cohort formed through a sequence of large-scale disruptions develops an involuntary and partial antifragility that constitutes a genuine epistemic advantage at a moment of transition.
Strauss-Howe generational archetypes cycle through history, each shaped by the phase of the Saeculum in which they came of age. The Xennial micro-cohort (born roughly 1977-1983) occupies a specific and unusually information-rich position in this cycle.
The Xennial cohort — born approximately 1977 to 1983 — is not a mere demographic curiosity. It is a specific epistemic formation. To understand why people born in this window can see what they can see, and why communicating it to younger people is structurally difficult, requires understanding what that formation consists of and why it produces the specific pattern recognition being experienced in February 2026.
The Formation: A Sequence of Epistemic Events
In childhood: the late Cold War, including its genuine existential tension. The nuclear dread of the early 1980s was not theatrical. Children born at that time absorbed it as lived atmosphere. Simultaneously: an analogue world of physical media, physical money, physical communication — experienced not as nostalgia but as functional normality, the baseline against which all subsequent change would be measured.
In early adolescence: the abrupt end of the Cold War. The 1989-1991 period was, for people aged nine to twelve, a formative experience of world-historical discontinuity. A wall fell. An empire dissolved. Things presented as permanent turned out not to be. This instals, at a foundational level, the conviction that apparently stable systems can end — not gradually, but suddenly, visibly, and completely. This is not a trivial epistemic event for a developing mind. It is a permanent update to the prior probability assigned to systemic stability.
In late adolescence and early adulthood: the internet arrived — not as something that had always been there, as it is for everyone born after approximately 1990, but as something that manifestly changed everything in real time. The Xennial cohort is the last generation to have clear memories of the pre-internet world as functional normality and the first generation to have adopted the internet as a central organising technology of adult life. This dual experience — analogue formation, digital adulthood — creates a specific cognitive advantage in the present moment: an internal reference point for what 'before' looks like that younger people, for whom the digital environment is simply the world, cannot access.
In young adulthood: the dotcom boom and bust, followed by 9/11, followed by the Iraq War. Three consecutive events each demonstrating that official narratives were unreliable and institutional authority could be catastrophically wrong. The WMD case was challenged not primarily by investigative journalism but by the civilian internet operating ahead of institutional media. The Xennial cohort processed this as a formative lesson about the epistemic gap between official accounts and distributed reality. Then 2008: confirming that the financial institutions presenting themselves as sophisticated managers of complex systems were either catastrophically incompetent or catastrophically dishonest or both. In Kahneman's terms, a System 1 trained on multiple high-quality feedback cycles of large-scale institutional failure.
Why Younger People Cannot Easily See This
The difficulty of communicating this pattern recognition to people born after approximately 1990 is not a failure of intelligence or education. It is a structural feature of epistemic formation. They did not experience the Cold War as tension. They did not experience the pre-internet world as normality. Most critically, they have not lived through enough complete cycles of the relevant kind to have trained their pattern-matching on the specific signature of this type of event. They are not missing data. They are missing feedback loops — the experience of having held a position, watched events unfold, and revised it based on how reality compared to expectation, repeated enough times to produce reliable intuition.
The educational and informational environment in which younger cohorts were formed was, during their formation, systematically optimistic about institutions. The period roughly 1995 to 2015 was one of genuine institutional performance — rising living standards, expanding rights, global poverty reduction, technological capability expanding into everyday life. People whose formative years fall there have a System 1 calibrated on institutional competence and legitimate authority as default. The current period feels to them like an anomaly rather than a pattern. They are looking for the return to normal. The Xennial observer is not looking for normal. They are watching the next phase of a cycle they have seen move before.
The most useful thing the Xennial generation can offer the younger ones is not conclusions. It is method. Not: this is what is happening. But: this is how you read a situation like this — these are the signals, this is what they mean, this is the historical record of what tends to follow. Walk them through a previous cycle not as received history but as lived sequence — what it felt like to be inside it, what the confident accounts looked like before the inflection point, what the warning signs were that were visible in advance to careful observers and dismissed by everyone else. The aim is not to transfer your conclusion but to help younger people develop the time-series intuition that your age and specific formation have given you access to.
This is not the first time in history that a combination of technological revolution and institutional crisis has produced the sensation of standing at an incomprehensible threshold. The parallels are instructive not because history repeats literally — it does not — but because the social and psychological patterns recur. Recognising the pattern is not the same as predicting the outcome. But it provides something more useful than prediction: a map of the forces likely to be in play and the phase transitions the map suggests lie ahead.
Gutenberg's press did not merely make books cheaper. It destroyed the Catholic Church's monopoly on textual interpretation — the ability to control who read what, in what language, with what gloss. Within eighty years, Luther's theses had spread across Europe at a speed previously impossible and thereafter unstoppable. The Church's response — the Index Librorum Prohibitorum, systematic deletion of the archive — was the fifteenth century equivalent of what we are watching now. It failed completely. What followed was not smooth transition to a better order but the Thirty Years War — decades of catastrophic violence as old power structures tried to contain something architecturally uncontainable. The technology was not benevolent. It was amplifying.
The industrial revolution is the closest historical analogue to the present moment: the first time technology produced genuine exponential economic growth. The immediate experience was not liberation. It was the destruction of existing social structures at a speed that support systems could not accommodate. The Luddites correctly identified that the technology was destroying their economic position. The framing of 'Luddite' as an insult is a retrospective judgement made by people who know how the story ended. Those living through it did not. The technology produced, eventually, historically unprecedented material abundance. It also produced the social dislocation that generated Marxism, the trade union movement, the welfare state — and the conditions for two world wars. Technology does not deliver utopia. It delivers amplified human nature, which contains everything.
The closest recent parallel — lived through in real time by the Xennial cohort. The dotcom bubble was rational exuberance about a genuinely transformative technology arriving faster than business models could accommodate. The serious disruption — to media, retail, political communication, the distribution of social authority — took another decade to fully manifest after the bubble collapsed. Those watching carefully in 1997 could see the shape of what was coming. Those around them could not, because the technology was still on the first half of the chessboard. The lesson: the early-stage observer of an exponential technology is not wrong simply because the majority cannot yet see what they are describing. They are ahead of the curve by the structural logic of exponential adoption at this point on the curve.
The pattern across all three: transformative technology arrives. Early adopters identify the implications. Existing power structures attempt suppression or containment. The attempt fails because the technology is architecturally uncontainable. A disruptive transition period follows, characterised by the specific social pathologies of interregnum. A new equilibrium eventually emerges that is qualitatively incomparable with the old one.
What is different about the present moment is the pace. The printing press took eighty years to destroy the Church's interpretive monopoly. The industrial revolution took thirty years to produce mass urbanisation. The internet took fifteen years to destroy the newspaper business model. AI is operating on a timeline of years. The institutional lag — the gap between technological reality and institutional adaptation — is proportionally larger. This is why the present moment feels more vertiginous. It is not because previous transitions were less significant. It is because this one is moving faster, and the gap between what is happening and what official accounts say is happening is correspondingly wider.
METR — Model Evaluation and Threat Research — published their Time Horizon 1.1 findings in January 2026, presenting longitudinal data on AI agent capability from 2019 to 2026. Unlike benchmark scores, which measure narrow performance on fixed tests and are subject to Goodhart's Law (the measure becomes the target and ceases to be a good measure), the time horizon metric measures something structural: how long a task can be before the agent fails at it. This resists gaming.
The overhang problem describes the gap between what technology can do and what economic and social systems have adapted to accommodate. When the overhang is large — when capability far exceeds adaptation — the eventual adaptation event is rapid and discontinuous rather than gradual. The METR data suggests the overhang is already very large and growing faster than adaptation.
There is a particular quality of clarity that comes from looking at a well-constructed graph of an exponential trend in real data. The METR time horizon study provides exactly that. What makes it significant in the context of this analysis is not merely that it shows AI improving — that is widely reported and widely contested. It is that it measures AI improvement using a metric that is structurally resistant to the usual objections, and that the resulting data has a shape that makes the theoretical arguments of the previous chapters empirically concrete.
The metric is elegant in its simplicity. Take the most capable AI agent available at a given point in time. Give it tasks of varying lengths — measured in the time a skilled human professional takes to complete them, ranging from under thirty seconds to over eight hours. Measure the success rate as a function of task length. Find the task length at which the agent succeeds fifty percent of the time. That is the time horizon. Repeat across multiple generations of models from 2019 to 2026. Plot the results. The time horizon has been doubling every seven months.
The R-squared of 0.83 between task length and agent success rate is the first important number. It tells you that the metric is measuring something real — that task length is a genuine predictor of difficulty for AI agents, not an artefact of the specific tasks chosen. The second important number is the doubling rate: seven months over the full period, four months in 2024-2025. The third important number is the current time horizon: over fourteen hours of human work, autonomously completable by frontier agents in early 2026. In late 2022, when ChatGPT launched, the equivalent figure was approximately thirty seconds.
What the Extrapolation Actually Means
The extrapolation from these numbers is not speculative in the way that most AI forecasting is speculative. It is mathematical projection of a measured trend. The caveats are real — the trend might slow, might plateau, might encounter obstacles not visible in the data. These are legitimate uncertainties and the METR researchers acknowledge them explicitly. But the default assumption, in the absence of a specific mechanism for why the trend would stop, is that it continues. And the METR researchers also acknowledge the reverse uncertainty: the trend might accelerate further, and they describe a plausible mechanism — AI-assisted AI development — that would produce exactly that acceleration.
A work-day of autonomous task completion means something specific. It means that an AI agent, given a well-specified problem of the kind that currently occupies a skilled professional for eight hours, will reliably produce an adequate solution without human intervention. This is not assistance. This is substitution at the level of the task. Extrapolate to work-week tasks: the agent can take a project brief on Monday and return a completed deliverable by Friday without human involvement in the intervening steps. Extrapolate to work-month tasks: entire project cycles — research, analysis, drafting, iteration, finalisation — become automatable in the same sense that manufacturing assembly became automatable in the twentieth century.
The sector-by-sector implications of this are uneven but comprehensive. Tasks that are primarily cognitive, well-specified, and output-measurable — the core of white-collar professional work — are precisely the tasks that the time horizon metric describes. Legal research. Financial analysis. Software development. Medical literature synthesis. Architectural design iteration. Journalistic research and drafting. These are the activities of the professional middle class, the activities that have historically provided both economic security and social identity to the largest educated segment of Western populations. The time horizon data describes the timeline on which those activities become automatable.
The Flywheel and the Event Horizon
The METR analysis identifies the mechanism that could push the trend from exponential to superexponential: AI-assisted AI research. As agents become capable of completing longer tasks autonomously, they become capable of completing longer AI research tasks autonomously. The researchers who build the next generation of models are themselves augmented by the current generation. The cycle shortens. The doubling rate compresses. The flywheel of acceleration that the singularity thesis describes as theoretical becomes, at some point, empirical — and the METR data suggests we are approaching that point rather than merely theorising about it.
This is the finding that connects most directly to the convergence thesis of this essay. The institutional systems that would need to manage this transition — regulatory frameworks, educational curricula, labour market structures, social safety nets, democratic deliberation processes — operate on timescales measured in years. The METR data describes a technology operating on timescales measured in months, with a doubling rate that may itself be compressing. The gap between these two clocks is not a policy problem to be solved with more agile governance. It is a structural mismatch between the pace of a technology and the pace of the human systems designed to accommodate technological change. That mismatch is the defining feature of the present moment, and the METR data makes it measurable rather than merely intuitive.
One additional implication deserves emphasis. The time horizon metric measures current frontier models. It does not measure what is in development. There is typically a gap of six to eighteen months between the training of a model and its public deployment. The capabilities that will be publicly visible in 2027 are being trained now. The capabilities that will shape 2028 are being architecturally designed now. The public experience of AI capability is always a lagged indicator of the actual frontier. The METR data, already striking, is a trailing measure. The leading edge is further along the curve than the published numbers describe.
Kahneman's System 1 processes information rapidly and associatively, generating confident conclusions from pattern-matching. Applied to exponential phenomena it fires wrongly — producing systematic underestimation at early stages and panic at late ones. The errors are not random. They are structured by the mismatch between the cognitive tool and the phenomenon.
Girard's mimetic theory holds that human desire is fundamentally imitative — we want what others want, not because of the object's intrinsic value but because others wanting it signals its value. Algorithmic social media, optimised for engagement, amplifies and accelerates mimetic contagion at a scale no previous medium achieved.
Turchin's structural-demographic theory identifies elite overproduction — the generation of more elite aspirants than elite positions can accommodate — as a primary driver of political instability. When the credential-to-position ratio inverts, frustrated elite aspirants become the primary destabilising force in a society.
The structural forces described in the preceding chapters — institutional decay, exponential technology, monetary transformation — do not operate in a vacuum. They operate through human beings with specific cognitive architectures, within social systems that amplify and distort individual responses into collective behaviour. Understanding how the transition will be experienced psychologically and socially is not secondary to understanding it structurally. In many respects it is more immediately consequential, because the transition will be navigated — or failed to be navigated — by actual human beings responding to actual events through the psychological machinery they possess.
The Cognition Problem: System 1 Meets the Exponential
The specific cognitive failure mode of the present moment is one Kahneman's framework identifies precisely. System 1 — the fast, automatic, associative processing system that handles most of daily cognition — applies linear extrapolation to most phenomena by default. This is a reasonable heuristic for most of the situations human beings have historically encountered. Most trends in a stable environment are approximately linear across short time horizons. The heuristic works.
Applied to exponential phenomena it fails systematically, and in a specific direction. At early stages of exponential growth, linear extrapolation produces underestimation — the curve looks flat because the early doubles are small in absolute terms. The person who dismissed the significance of the internet in 1996 was not unintelligent. Their System 1 was applying the standard heuristic to a phenomenon that violated its assumptions. At later stages, the same heuristic produces panic — the sudden recognition that the trajectory was non-linear triggers a recalibration that overshoots, because System 1 cannot smoothly update to the correct mathematical model. It oscillates between dismissal and alarm. Neither is calibrated. Both are highly confident.
This is the psychological structure of the present moment at the population level. A large majority of people are in the early-stage underestimation phase — the METR data, the AlphaFold result, the doubling rates, register as impressive but not threatening, because linear extrapolation from current capability produces a manageable near-term picture. A smaller, growing minority has made the cognitive transition to exponential models and is in various states of alarm, or strategic repositioning, or quiet preparation. The gap between these two populations — in their sense of urgency, their interpretation of current events, their behaviour — is one of the defining social fractures of the moment.
Girard's Machine: Mimetic Desire at Scale
René Girard's insight that human desire is fundamentally imitative was developed in literary analysis but extends with uncomfortable precision to social media dynamics. The core mechanism: we learn what to want by watching what others want. In a world of information scarcity and slow communication, mimetic contagion was constrained by the speed at which desire-signals could propagate. The algorithm changes this entirely. Social media platforms, optimised for engagement, discovered empirically what Girard described theoretically: that mimetic signals — what is desired, feared, admired, hated — generate the most reliable engagement, and that amplifying them maximises the metric that matters to the platform's business model.
The result is a global nervous system for human desire, operating in real time, optimised for contagion rather than accuracy. Political movements, financial manias, moral panics, cultural trends — all propagate faster, reach further, and intensify more rapidly than was structurally possible before this architecture existed. This is not neutral amplification. It is selective amplification of the most mimetically potent signals, which are disproportionately those that trigger social comparison, status anxiety, and outgroup differentiation. The political polarisation, the epistemically fragmented information environment, the oscillation between collective euphoria and collective panic — these are not pathologies of the system. They are the outputs of a system operating correctly according to its actual design criteria.
Le Bon's Crowd, Turchin's Pressure
Gustave Le Bon, writing in 1895, described the psychology of crowds with an accuracy that reads as prophetic when applied to algorithmically-mediated collective behaviour. The crowd, he observed, does not reason. It feels. It is susceptible to suggestion and contagion in ways that individuals alone are not. It simplifies, it polarises, it responds to symbols and narratives rather than evidence and argument. The crowd rewards those who speak with emotional certainty and punishes those who express calibrated uncertainty. Le Bon was describing pre-digital street crowds. He was describing 2026 in structural terms that require only scale substitution.
Peter Turchin's structural-demographic theory adds the supply-side pressure that transforms mimetic contagion from a background feature of social life into a destabilising force. Elite overproduction — the systematic generation of more credentialled aspirants than elite positions can accommodate — creates a pool of frustrated, articulate, ideologically motivated individuals who have been promised access to status they cannot obtain through the conventional channels. These are not the dispossessed and marginalised. They are the over-educated and under-positioned. They have the skills to organise, communicate, and lead movements. They have the grievances to motivate them. They are the available fuel for the fire that Girard's machine provides the ignition for.
The combination is precise: exponential technology producing rapid disruption that System 1 cannot process accurately; mimetic amplification of the anxiety and confusion this produces; elite overproduction generating the motivated actors who channel collective emotion into political movements. The institutions designed to manage social stress — deliberative democracy, professional media, academic expertise — are themselves the institutions losing legitimacy fastest. The stabilising mechanisms are weakening at exactly the moment the destabilising pressures are intensifying.
The appropriate response to this analysis is not despair. It is calibration. The person who understands these dynamics is not above them — everyone operates in the same information environment, through the same cognitive architecture. But understanding the mechanism creates the possibility of compensating for it: deliberately applying System 2 to exponential phenomena; curating an information environment less optimised for mimetic contagion; maintaining connections to people who are differently positioned in the epistemic landscape. This is what psychological preparation for the transition actually means at the cognitive level.
The Stoic practice of premeditatio malorum — premeditation of adversity — is not pessimism. It is the deliberate mental rehearsal of difficult futures in order to reduce their shock and increase the quality of one's response when they arrive. Seneca: 'Omnia, Lucili, aliena sunt, tempus tantum nostrum est.' What belongs to us is not outcomes but our orientation toward them.
Taleb's barbell strategy in investing — maximum safety at one end, asymmetric upside at the other, nothing in the middle — generalises to life strategy under radical uncertainty. The worst position is the middle: moderate risk, moderate return, maximum fragility to tail events.
Frankl's logotherapy holds that meaning — not pleasure, not power — is the primary human motivational force, and that meaning can be found in suffering as much as in success. Periods of radical disruption, which destroy existing sources of meaning, require the construction of new ones. This is active work, not passive recovery.
Elinor Ostrom's work on collective action demonstrates that communities can manage shared resources and shared challenges without central coordination, provided they develop appropriate institutional arrangements. The relevant unit of preparation is not only the individual but the community.
This chapter differs from the preceding ones in orientation. The previous chapters have been analytical — attempting to map what is happening and why, using frameworks that illuminate the structure of the present moment. This chapter is practical. Given that the analysis is substantially correct, given that the METR data is a reasonable description of the trajectory of AI capability, given that the fiat order is under structural stress and institutional legitimacy is eroding — what does a thoughtful person actually do? What does preparation look like when the thing you are preparing for is not a single identifiable event but a rapid, multi-domain transition whose specific form is genuinely uncertain?
The answer is not a checklist. Checklists are appropriate for known risks with defined responses. What we are facing is a known uncertainty — we know the direction of travel with high confidence, but the specific form, timing, and sequence of disruptions is genuinely unclear. The appropriate response to known uncertainty is not preparation for a specific scenario but the cultivation of what might be called adaptive capacity: the combination of material, social, psychological, and epistemic resources that make you resilient across a wide range of possible futures rather than optimally positioned for one specific one.
Financial Preparation: The Barbell in Practice
Taleb's barbell strategy, developed in the context of investment portfolios, generalises powerfully to financial preparation under radical uncertainty. The core insight is that the worst position is the one that feels safest: moderate risk, moderate return, maximum exposure to tail events. A salary with standard savings in conventional assets is exactly this position — it offers no protection against the scenarios that matter (sustained high inflation, sector-wide employment disruption, institutional financial instability) while providing no asymmetric upside from the scenarios that could be transformative.
The barbell applied to personal finance in the present moment has a specific structure. On the safety end: reduce dependence on single income sources, maintain genuine liquidity (not assets that feel liquid until you need to sell them in a crisis), and hold some portion of savings outside the conventional financial system — whether in physical assets, commodities, or in hard monetary assets with the properties described in Chapter IV. The point is not that conventional financial assets will fail. It is that the correlation between conventional assets increases dramatically in tail scenarios, making apparent diversification illusory precisely when diversification is most needed.
On the upside end of the barbell: allocate a portion of capital and, more importantly, attention and skill-building to asymmetric opportunities created by the transition itself. The industrial revolution destroyed handloom weavers and created factory owners, engineers, and eventually an entirely new professional class. The internet destroyed newspaper classified advertising and created the entire digital economy. Each transition destroys value in the existing structure and creates it in the new one. The question is positioning: are you holding assets and skills that are primarily valuable in the old structure, or ones that are valuable in the new one, or ideally ones that are valuable in the transition itself?
Professional Preparation: Skills for the Transition
The METR time horizon data describes which skills become automatable and on what timeline. The skills most immediately at risk are those that are primarily cognitive, well-specified, and output-measurable — the core of white-collar professional competence. This does not mean those professions disappear overnight. It means they transform: the ratio of human judgment to routine cognitive execution shifts, the number of people required to produce a given output falls, and the skills that command premium compensation change.
The skills that remain valuable — and in many cases become more valuable — through this transition share certain characteristics. They involve judgment in genuinely novel situations without clear right answers. They involve relationship and trust in contexts where those cannot be delegated to a system. They involve the ability to direct and evaluate AI outputs effectively — which is itself a skill that must be developed, not a natural extension of existing professional competence. And they involve what might be called epistemic positioning: the ability to operate effectively in conditions of high uncertainty, to update rapidly on new information, and to avoid the false comfort of frameworks that no longer fit the world.
Practically, this means that professional preparation in the present moment has a specific character. Develop deep familiarity with AI tools as they exist now — not as a user of surface features but as someone who understands what these systems can and cannot do, where they fail, how to specify tasks for them effectively, and how to evaluate their outputs critically. This fluency is the closest thing to a universally portable skill in the transition period, because every profession will need people who can navigate the human-AI interface competently. The person who develops this fluency early is positioned to be the navigator rather than the navigated.
Equally important: maintain and develop skills and relationships that cannot be automated. Deep domain expertise that involves genuine judgment in ambiguous situations. The ability to build trust with other humans in high-stakes contexts. Physical and craft skills that depend on embodied knowledge and presence. Creative work that is valued for its human origin rather than merely its output. None of these are immune to disruption in the long run — the METR extrapolation eventually covers them. But they are more durable in the medium run, and the medium run is the planning horizon that matters for most people making decisions now.
Social Preparation: Community as Infrastructure
The most underrated dimension of preparation for large-scale societal disruption is social: the quality and density of your human relationships and community ties. This is not a soft observation. It is empirically grounded in the literature on how communities survive and recover from shocks. Ostrom's research on collective action demonstrates that the communities that manage shared challenges most effectively are not those with the most resources or the most central coordination, but those with the strongest internal trust networks, the clearest shared norms, and the most developed capacity for mutual support.
The individualism that has characterised the social organisation of the neoliberal period — the reduction of social ties to market transactions, the atrophying of community institutions, the privatisation of meaning into consumption — represents a form of social fragility that becomes acutely visible under stress. A person with strong community ties, mutual obligations, and a network of trusted relationships has resources available in a crisis that no amount of individual financial preparation can substitute for. The rebuilding of those ties — in whatever form is authentic to your context — is not a nice-to-have. It is infrastructure.
Practically, this means investing time and energy in relationships and communities that are not purely transactional. Local relationships — with neighbours, with local institutions, with people who share physical space rather than only digital space — have a resilience that purely digital networks lack. Skill-sharing networks, community gardens, local mutual aid structures, religious or civic communities, informal networks of trust and reciprocity: these are not nostalgia. They are the appropriate social infrastructure for a period of rapid change and institutional stress.
Psychological Preparation: Groundedness Under Uncertainty
The psychological challenge of the present moment is specific and worth naming precisely. It is not anxiety about a single defined threat — that is manageable, because the threat is bounded and the response can be calibrated. It is the challenge of sustained orientation under conditions of genuine, multi-domain, open-ended uncertainty. The institutions that normally provide psychological ground — stable employment, functioning governance, trustworthy financial systems, coherent cultural narrative — are all under simultaneous pressure. The scaffolding on which most people build their sense of stability is visibly weakening.
The Stoic tradition, specifically the practice of premeditatio malorum, offers something genuinely useful here. The practice is not catastrophising. It is the deliberate, calm contemplation of adverse futures in order to reduce their psychological impact and improve the quality of one's response when they arrive. The person who has mentally rehearsed the possibility of significant employment disruption, who has thought through what they would do and who they would turn to, responds to that disruption differently than the person for whom it arrives as a complete shock. Not better in every dimension — the Stoics did not promise the elimination of suffering. But with more equanimity and more effective action.
Frankl's insight from the extreme laboratory of Auschwitz is relevant here in a way that is not diminished by the difference in scale. The people who survived psychologically intact — not merely physically, which was largely beyond their control — were those who were able to maintain a sense of meaning that was not contingent on their circumstances. Meaning that was located in relationships, in contribution, in the act of bearing witness, in the commitment to something beyond survival itself. In a period of disruption that destroys existing sources of meaning — job identity, institutional loyalty, the narrative of continuous progress — the active construction of meaning is not optional. It is the psychological work of the transition.
The Deeper Preparation: Epistemics and Adaptability
Beneath the practical dimensions of preparation — financial, professional, social, psychological — there is a more fundamental one: the quality of your model of reality. All the practical preparations are only as good as the understanding that guides them. A person with an accurate model of what is happening and why will make better decisions across all the practical dimensions, even with fewer resources, than a person with more resources and an inaccurate model.
Epistemic preparation means several things. It means developing the habit of updating your beliefs on evidence rather than maintaining them for comfort. It means seeking out sources and interlocutors who will challenge your model rather than confirm it. It means distinguishing between conclusions — which should be held with appropriate tentativeness — and the methodology that generates them — which should be robust and consistently applied. And it means maintaining calibrated uncertainty: neither the paralysis of excessive doubt nor the false confidence of a model that cannot accommodate surprise.
Finally: hold the analysis lightly. This essay has argued a specific position with analytical rigour, because rigour is more useful than vagueness. But the position is held with genuine uncertainty, and the appropriate response to genuine uncertainty is not the abandonment of analysis but the maintenance of the capacity to revise it. The world is moving fast. The frameworks that are most useful today may need updating in six months. The preparation is not to find the right answer and hold it tightly. It is to develop the capacity to find better answers continuously, faster than the ground shifts beneath them.
Hobbes conceived the Leviathan — the centralised sovereign — as the solution to the war of all against all. The algorithmic equivalent is a state or corporate infrastructure with complete visibility into individual behaviour, the capacity to score and rank it, and the ability to adjust access to social goods accordingly. The enforcement mechanism is not violence but access control.
Ostrom's work on commons governance demonstrates that decentralised, polycentric systems with appropriate rules can manage shared resources effectively without central authority. The Sovereign Mesh is this applied to cognitive infrastructure: distributed AI capability governed by local and community-level rules rather than centralised platforms.
Hayek's insight that distributed knowledge cannot be aggregated by any central authority points to the epistemic vulnerability of centralised AI systems: they encode the biases and errors of their designers at scale, whereas distributed systems allow for error correction through competition and diversity.
Every major technological transition produces an architectural fork — a point at which the infrastructure being built will determine, for generations, the distribution of power, privacy, and agency within the society that builds on it. The printing press forked between censored and uncensored information flows. The industrial revolution forked between company towns and free labour markets. The early internet forked between open protocols and proprietary platforms. In each case, the choices made during the architectural phase proved enormously durable — the structures built on top of infrastructure inherit its properties.
We are at the architectural fork for AI. The decisions being made now — about where computation happens, who controls the models, what data is retained, how outputs are governed — are establishing the infrastructure on which everything that follows will be built. Two trajectories are currently live, and the gap between them is not a matter of technical preference but of civilisational structure.
The Algorithmic Leviathan
In the first trajectory, the AI capability curve is captured by a small number of large institutions — states and corporations, often in alliance — who use it to extend surveillance, scoring, and behavioural management to previously unmanageable scales. The mechanism is already visible in partial form. Predictive systems inform credit decisions, insurance pricing, hiring, and policing. Social credit architectures, pioneered in specific national contexts, provide the template. The convergence of AI capability with existing institutional incentives to monitor and manage populations produces a system in which individual behaviour is continuously scored, the scores determine access to social goods, and the scoring criteria are opaque and unappealable.
This is not science fiction. It is the logical extension of currently deployed systems, operating on the capability trajectory the METR data describes. A system that can complete work-month tasks autonomously can also monitor, analyse, and score behaviour at scale without human review. The computational cost of comprehensive surveillance drops toward zero as the capability curve rises. The Algorithmic Leviathan does not require malign intent. It requires only the operation of institutional self-interest — the same mechanism that produced the Tainter dynamic — applied to the infrastructure of AI.
The Sovereign Mesh
In the second trajectory, the distribution of AI capability follows a different path. Models run locally, on hardware owned by individuals and communities. The capability gains from the exponential curve are distributed rather than captured. A small team with AI fluency and local models competes effectively with large organisations that have more headcount but less capability per person. The information asymmetry between institutions and individuals narrows rather than widens, because the tools of analysis and verification become available to anyone rather than exclusively to those with institutional access.
The Sovereign Mesh is not utopian. It has its own failure modes — the same AI capability that empowers individuals also empowers malicious actors, and the absence of central coordination creates coordination problems that well-functioning institutions solve efficiently. Ostrom's lesson is relevant here: decentralised systems work when they develop appropriate governance structures, and those structures require active construction and maintenance. The mesh does not self-organise into a functioning commons. It requires the kind of deliberate institutional design that Ostrom's work illuminates.
Why the Fork Matters Now
The reason this analysis belongs in a chapter of its own is that the architectural fork is not a future choice. It is a present one, and it is being made continuously, by the aggregated decisions of developers, investors, regulators, and users. The open-source AI movement, the development of local inference capabilities, the resistance to centralised model control — these are not merely technical preferences. They are architectural choices about which trajectory the capability curve serves.
The convergence thesis suggests that the institutional decay documented in Chapter I, the monetary dynamics of Chapter IV, and the capability trajectory of Chapters III and VII are all producing conditions in which the architectural fork becomes increasingly consequential. A world of weakening institutional legitimacy, eroding monetary privacy, and rapidly expanding AI capability is a world in which the question of who controls the AI infrastructure is not peripheral. It is the question that subsumes most others. The person who understands the fork, and who makes choices — personal, professional, political — that push toward the Sovereign Mesh rather than the Algorithmic Leviathan, is doing something more consequential than most of what currently passes for political engagement.
The full analytical picture that emerges from applying these frameworks simultaneously to the present moment is coherent, internally consistent, and in several respects deeply uncomfortable. It does not require the existence of any coordinating malign intelligence. It requires only the operation of well-understood mechanisms — institutional self-interest, exponential technology, monetary dynamics, generational epistemology — at a specific historical juncture where they are all operating simultaneously and in the same direction.
The institutions of the Western order are exhibiting the classic signature of Tainter systems approaching critical transition: increasing complexity with diminishing returns, increasing variance in outputs, increasing autocorrelation, lengthening recovery times, and the deployment of the tools of accountability to protect institutions from accountability. The deletion of criminal archives. The management of economic statistics. The selective application of legal process. These are not signs of evil. They are signs of systems doing what systems do at this point in the Tainter cycle.
Simultaneously, the technology of transparency is destroying the information asymmetry on which institutional power has always depended — not gradually but exponentially, at a rate now faster than institutional response time. For the first time in recorded history, the tools available to a person who wants to know something are more powerful than the tools available to an institution that wants to suppress something. This reversal is producing exactly the institutional panic and increasingly clumsy self-protective behaviour one would predict.
The METR time horizon data makes the technological acceleration empirically concrete in a way that intuition and extrapolation cannot. A measured doubling every seven months — accelerating to every four months in the most recent period — describes a trajectory on which autonomous AI agents complete work-day tasks in 2027, work-week tasks in 2028, and work-month tasks in 2029. These are not predictions derived from theoretical models. They are projections of measured trends. The R-squared of 0.83 in the underlying data is the kind of signal that, in any other domain, would be considered a strong empirical finding requiring serious policy response. In the domain of AI capability, it is largely absent from mainstream institutional discourse.
The monetary system underlying the political order is encountering its logical antithesis in Bitcoin — a technology that reproduces the key properties of sound money in digital form without depending on institutional trust. Whether Bitcoin specifically is the monetary technology of the next paradigm remains uncertain and is perhaps the wrong question. The right question is whether the fiat architecture can maintain the institutional credibility it requires as its informational scaffolding is systematically removed. The answer, across a horizon of years, appears to be no. The free market has identified this gap and is filling it with a technology designed for the world that is emerging.
The preparation chapter attempts something harder than analysis: it tries to translate the structural understanding into practical orientation. The barbell approach to financial resilience. The development of skills that navigate the human-AI interface rather than competing against it. The rebuilding of community infrastructure that individualism has allowed to atrophy. The psychological practice of equanimity under genuine uncertainty. And beneath all of it, the epistemic work of maintaining an accurate, updateable model of what is actually happening rather than the comforting model that says the rate of change will eventually slow to something manageable.
What comes next is genuinely uncertain in its specific form, though clearer in its structural dynamics. The transition will be disruptive before it is liberating. The printing press gave us the Reformation before it gave us the Enlightenment, and the Thirty Years War in between. Transformative technology does not deliver utopia. It delivers amplified human nature — which contains everything, the best and worst of what we are capable of. The outcome depends on the quality of the navigation, on whether the people making decisions during the transition can apply sufficiently accurate models of what is happening to make choices that lead toward the better possibilities rather than the worse ones.
