By Stephen F. DeAngelis.
A Reader’s Compass
A virus that began in one province was a planetary fact inside a season, and the bodies built to answer it were reading the morning brief while the thing crossed three borders. The inherited instruments are doing the work they were built to do, at the cadence and the resolution they were built to do it at. The work no longer reaches the problem, because the problem has moved to a cadence and a resolution the instruments were never tuned to catch.
What governance produces, when it is governing, is decisions a serious actor can defend in front of the consequences they produced. Decisions like that rest on knowledge that has been vouched for as true (validation), seen at the resolution where the problem lives (grain), and delivered in time to act (clock). I am not using “coherent” in the epistemologist’s sense of beliefs that hang together; I mean knowledge that has earned the right to support a decision by satisfying three external tests. Validation is the heaviest of the three because the other two depend on it. The inherited sovereign was built where those three components held by default. The conditions moved.
If you cannot grasp a problem in its real context, at the pace it unfolds, and with reliable facts, you are not truly in control of it. At best, you are governing a simplified picture and mistaking the picture for the world.
I.
Three instruments carry the weight of what we call sovereignty, and they were built in different centuries to answer different scarcities. The Westphalian state made territorial jurisdiction the unit of legitimate authority. The corporate form made the firm an actor whose decisions answered to itself. The multilateral body gave the state a forum in which problems too large for any one state could be worked. Each was so successful at the answer it was built to provide that we stopped noticing it was an answer to anything and began treating it as the natural shape of authority. Under the conditions of its century, each instrument produced coherent knowledge by default, and the three components of that knowledge held without effort. It was the conditions that had moved.
Begin with the oldest. Westphalia fixed a unit. Inside a territory, one authority. The wars that preceded it had been fought across borders that meant nothing to the armies crossing them, over a confessional authority that claimed jurisdiction everywhere and therefore held it nowhere. The settlement organized European order around territorial jurisdiction for three and a half centuries, and it organized it well, because the problems that needed solving moved at the speed of armies and trade, of news carried by the post, at a resolution a bounded territory could read and on a cadence a court could keep. Inside that bounded jurisdiction, the validation of public claims (the church’s claim to authority, the prince’s claim to law, the magistrate’s claim to fact) could be performed at a pace and a resolution the institutions of the territory could sustain.
The corporate form arrived later and answered a different scarcity. The nineteenth century scaled the joint-stock company past what any set of partners could pledge. General incorporation laws turned the chartered exception into a thing any group of subscribers could call into being, and limited liability made bounded risk a standing feature rather than a privilege the sovereign granted one venture at a time. The corporation now organized the industrial economy around an actor whose authority over its own choices matched the cadence of the operating environment it lived in. A firm could read its market at the resolution of the order book and the annual report, vouch for what it saw through audit and ledger, and act at the speed those documents arrived, which for a long time was the speed at which the market itself moved.
Within living memory, a third instrument was built by people who had watched the absence of such coordination cost the twentieth century its first half. The American historian Zachary Carter’s account of John Maynard Keynes across the Bretton Woods summer is the one I return to, because Carter holds the thing legible that the technical histories let slip.1 What came out of that summer was a forum: a place where the nation-state kept its authority and lent a portion of it to a structure that could coordinate what no single state could coordinate alone. The design conceded something the older two instruments had not. It admitted that a class of problems existed that no bounded jurisdiction could hold on its own, and it built a place where the bounded jurisdictions could stand together long enough to work them. For the conditions of its century that was enough, because the problems it was built to coordinate still moved at the cadence a conference of finance ministers could keep and lived at a resolution a national treasury could read. The achievement belonged to the apparatus, but the apparatus belonged to the people who had built it. Keynes’ own mind was the instrument that could see at the depth the problem required, reading the apparatus he had built at the resolution it actually moved at. That kind of ascent is the move the present has not yet managed.
Then the conditions moved in a direction none of the three was built to face. A pandemic is local and planetary in the same week. A credit shock crosses three time zones before a finance ministry has finished its morning brief. A climate signal arrives on a horizon longer than any electoral mandate and shorter than any constitutional revision can reach. A single piece of generated content can reach tens of millions of feeds within a day and hundreds of millions across the following days, an order of distribution the platform era has documented across viral image generation, deepfake video propagation, and synthetic political content in the 2024 and 2025 election cycles.2 None of these is a failure of the state, the firm, or the multilateral body at the work each was built for. Each is an observation about a world the three were not built to read, arriving at a speed and resolution none of them was tuned to, and arriving through a validation architecture that the present information environment has overrun.
The shape of the trouble is easy to miss because each instrument, taken alone, still looks like it is working. The state still passes laws, enforces them inside its borders, and collects the taxes that fund the enforcement. The firm still books revenue, files its statements, and answers to a board that reads them. The multilateral body still convenes, negotiates, and issues the communiqué at the end. Watch any one of them at the work it was built for and you will see an instrument in good repair. The failure shows only at the level of the problem, which no longer sits inside any one instrument’s frame. It arrives at each of the three as a partial thing, a slice the instrument can read, while the part that crossed into the next instrument’s territory, or fell below the resolution of all three, goes unread by every one of them. Each of the three is still doing the work it was built to do. The problem that crossed between them has no instrument tuned to read it.
Whether the inherited instruments can be evolved to meet the new question, or must be stood alongside instruments no one had imagined when they were built, is the open thing. The world the inherited sovereign was built for is not the world it now occupies.
II.
Before an instrument can act, it has to read. It takes in the environment at some level of detail and returns a picture, and everything the instrument decides afterward is decided about the picture rather than about the environment. A reading that drops the structure of the problem before the first question is asked has already conceded the decision. Grain is the resolution at which the reading happens; the inherited sovereign reads its world at the grain of jurisdiction, and the decisions now moving that world live at resolutions both finer and coarser than any single jurisdiction can hold.
Take the state. It was built to read at the resolution of bounded territory, and it reads transnational decisions the only way it can, through the proxy of national statistics that have already aggregated the structure out of the data before the analysis begins. A trade flow, a capital movement, a migration, a contagion: each lives at a scale that crosses the border and recombines on the far side, and the state meets it as a national number, a figure that records the territory’s slice of a thing whose shape is not territorial at all. The American political theorist Nils Gilman has named the broad form of this: the nation-state is structurally mismatched in scale to the planetary problem, too small for it and too large for the local one.3 The finer point is that the state does not merely sit at the wrong scale. It reads at a resolution that has dropped the structure of the problem before the first question is asked. A clean national figure returned about a transnational thing is a reading of a different thing altogether, one that has borrowed the first one’s name.
One level down, the firm carries the same mismatch. A board reads a chart with four categories about a recommendation that searched a space of four hundred entangled variables. The chart is not wrong so much as it is smaller than the decision it has been convened to govern, and the part of the decision that did not survive the trip up to the board is the part where the firm is most exposed. The form is working as designed, summarizing the operating reality into something a board can hold in a meeting.
What changed is the reality, not the board. The thing being summarized acquired a structure too fine to survive the summary, and the board now governs a picture that has had the deciding detail aggregated out of it before the picture reached the room.
The multilateral body inherits the loss twice. Built to read at the resolution of nation-state representation, it takes the planetary problem in through aggregated national positions, and those positions already carry the loss of the national instruments that produced them before the multilateral conversation has said its first word. A planetary signal enters the forum already twice-reduced, once by the territorial scale of each state’s reading and once by the diplomatic scale of each state’s position. The body then negotiates among positions at a scale that has nothing left in it of the planetary thing, and the communiqué that emerges is a careful settlement of a question that bears the planetary problem’s name while containing almost none of its actual shape. The forum is not failing to negotiate. It is negotiating skillfully over an artifact two reductions removed from the thing it was convened to address.
What the multilateral body stages at planetary scale arrives, at smaller scales, inside single rooms. I once sat in the office of a very senior government official in a country I was supporting on its own economic development, the morning an unsolicited offer from a competing nation-state arrived on his desk, a technological capability with hidden strings, a Trojan Horse. The junior official walked through the offer. When I named what it was, the incredulity on the senior official’s face was unmistakable; he decided in our favor that day. What had actually walked into the room was not one offer against another. It was several national positions pressing on one decision at once, and the official across the table was reading all of them through the single narrow resolution his office was built to read. The forum’s planetary failure compressed into one desk and one afternoon: a decision resolved at a scale far coarser than the problem underneath it.
The American philosopher John Dewey saw the human version of this a century ago. The public, he wrote in 1927, “seems to be lost; it is certainly bewildered,” and it is lost because the consequences of the transactions that bind it have grown too indirect, too far-flung, too tangled to trace back to the acts that produced them.4 The citizens have not lost the capacity to judge. They have lost the ability to see what they would be judging. Carry the diagnosis up to the sovereign and it transfers whole. A sovereign that cannot find the problem at the resolution where the problem lives cannot recognize what it has been convened to govern. A diagnosis taken at the wrong scale becomes a confident diagnosis of the wrong thing, and the confidence is what keeps anyone from noticing.
III.
Acting has its own clock, and the cadence determines whether the reading is still knowledge by the time the actor reaches the decision. A correction that arrives at the right resolution but at the wrong moment governs nothing, because the moment it was meant to govern has already passed. The inherited sovereign acts at the cadence of legislative deliberation, of the board cycle, of the negotiated treaty, and the decisions now moving the operating environment propagate at the cadence of the network. When the reading arrives after the moment of action has passed, the reading is a record of what the decision was made without.
The engineer’s vocabulary makes the asymmetry exact. Control theory says plainly that a system that responds more slowly than its environment changes is a system whose corrections always arrive after the change they were meant to correct. A correction that arrives late, into a system that has already moved on, can push in the wrong direction entirely and amplify the thing it was meant to dampen. A slow controller becomes a source of the instability it was installed to prevent. The mathematics belongs in a piece about sovereignty because sovereignty is, at the level of its operations, a control problem: an actor steering a system on the basis of what it can read about the system’s state.
Watch the state at planetary cadence. A pandemic propagates across continents in something like nine weeks. The information about it reaches every capital in nine seconds. The legislative response, built to the cadence of committee, hearing, markup, and ratified instrument, takes nine months on a good year.5 Set those against each other and the ratio between the speed of the threat and the speed of the answer is roughly that between seconds and months, a gap of seven orders of magnitude between the cadence of the information environment and the cadence of the legislative response, and several orders of magnitude between the cadence of the threat itself and that response. The state was not built for this. It was built for the cadence of bounded threats moving at the speed of armies and trade, where nine months was fast. The instrument that was fast enough for that world is structurally too slow for this one, by a margin no amount of diligence inside the old clock can close.
At machine cadence, the firm runs the same mismatch. A funding shock at one wholesale desk in London reaches dozens of institutions across four continents before the New York close, and the committee that would govern the propagation meets four times a year. The board’s authority was never just procedural. It was the place where the firm’s judgment was supposed to be exercised, the room where a human deliberation stood behind the decision and could be held to account for it. When the decision moves faster than the deliberation can convene, the firm does not simply act late. It acts without the deliberation having happened at all.
The form is not idle. There is a different version of the firm that can move at the right speed. When I model this with the people who run such firms, the gap is not abstract. The global marketplace can be modeled dynamically, and the right instrument applied to marketplace data lets a company read its market and its competitors at machine cadence. I have been in rooms where senior leaders acted decisively off such readings, because the instrument and the resolution and the data were right. The firm that does not move at that speed is deliberating at exactly the speed it was built to deliberate at, which is no longer a speed that catches the thing it is deliberating about, and a deliberation that arrives after the decision it was meant to govern is only the appearance of authority, laid over a decision that authority never touched.
Climate bends the same problem into a worse shape. On a horizon of decades, the signal arrives at a body that meets once a year. A decadal horizon read against an annual cycle is not, on its own, an impossible mismatch. It becomes impossible when the signal carries thresholds that, once crossed, do not reverse on any horizon shorter than centuries. The annual cycle is calibrated to a world in which most of what a body gets wrong this year it can correct next year. A body deliberating on an annual cycle about an irreversible threshold is working from the assumption that there will be a next year in which to correct, and for the thresholds that matter most, there is not.
The twentieth-century German political philosopher Hannah Arendt drew the line this widening gap falls across: between politics and administration, between the act and the management of acts.6 The cadence mismatch turns her distinction from a philosophical claim into a structural fact. The inherited sovereign can still administer. What it can no longer do is govern at the speed the environment now moves, and administration carried out at a cadence that cannot catch its environment is the exercise of power without the authority that would come from actually steering the thing.
IV.
Underneath grain and clock sits the component the inherited sovereign is losing most completely. Validation is the warrant that what an actor sees can be vouched for as true, and the warrant is what makes a reading available to a decision at all. The other two components depend on it. A reading at the wrong scale still tells an actor what they are looking at; a reading at the wrong moment still carries forward to inform what they prepare for next; neither failure removes the ground from under the decision. A reading without warrant does. The actor is not deciding badly. The actor is deciding without the conditions under which a decision is possible at all.
The mechanism of the failure is a cost asymmetry. The instruments the inherited sovereign used to validate public information were every one of them calibrated to a world in which producing plausible content was costly. The cost of production was the validation gate, doing its work silently, the way the cost of forging a document used to be the thing that made most documents trustworthy. Social media and generative systems together drove the cost of producing plausible content to something near zero while the cost of validating that content stayed exactly where it had always been. The gate was never defeated by an adversary. It dissolved, because the thing it was made of was a cost that no longer exists.
The asymmetry compounds in a way the raw cost figure understates. Producing a plausible falsehood is now nearly free, nearly instantaneous, and nearly infinitely reproducible, while every one of those three properties runs in the opposite direction for the act of checking. Validation is costly, it is slow, and it does not reproduce, because each instance of judgment has to be performed again by a mind that has actually attended to the thing. The resource that checking consumes is the finite attention of judging minds, the one resource no economy of scale has ever managed to expand.
The visible instance of this is the Gen AI systems that institutions across the public and private sectors are now deploying at scale. The systems produce plausible content. They cannot vouch for the truth of what they produce. They hallucinate facts the producer cannot distinguish from facts the world contains. They exercise no subtle human judgment because they have none to exercise; what they have is a statistical regularity in language that runs on a clock the producer cannot inspect at the resolution the user can audit. The black box is not a metaphor. It is the operating condition of the tools that are increasingly the instruments through which knowledge enters the polity. This is not a marginal observation. It is the convergent finding of the model evaluations the foundation-model laboratories themselves publish, the academic interpretability literature on what large language models actually do, and the regulatory architecture taking shape in the European Union, the United States, and the United Kingdom around outputs the regulators cannot inspect.7 The instruments are doing what they were built to do. The architecture they sit inside has not been built. A system that produces plausible content at machine speed without a human-cadence layer that can vouch for the content is producing knowledge that cannot support coherent decisions. The decisions still get made on it. They are not decisions in any sense the polity was built to make.
The German philosopher Jürgen Habermas saw the shape of this before the machinery existed to complete it. The bourgeois public sphere, he showed in 1962, was a historical achievement, not a permanent feature of the world. Inside its specific conditions, private people came together as a public and validated what passed in the polity by reasoning about it in front of one another, and the reasoning, not the reasoner’s rank, carried the day. Then the sphere was converted by the interpenetration of state and economy and the industrialization of mass media into a sphere that hosts something else. His sentence for what it became is the one the present moment keeps proving: “Publicity loses its critical function in favor of a staged display, even arguments are transmuted into symbols to which again one can not respond by arguing but only by identifying with them.”8 That is the whole of it, written about radio and the illustrated press, describing a machinery that would not be built for another half century. An argument you can answer by reasoning is a validation instrument. A symbol you can answer only by identifying with it or refusing it replaces validation. The current information environment is that conversion finished at machine speed, the staged display now generated and distributed faster than any public could convene to answer it.
The Czech dissident and former president Václav Havel watched the same failure from inside it. Havel found his image in a greengrocer who places in his shop window, among the onions and carrots, a sign carrying an official slogan he neither believes nor examines. The sign does not assert anything the greengrocer would defend if asked. It is the separation of the assertion from any cost of holding it, the public gesture emptied of the judgment that would make it a validation of anything, and Havel saw that a whole society could be organized around that separation without a single person having decided to lie. The machinery that now drives the cost of the public gesture to zero has detached the assertion from the judgment at a scale Havel was describing in miniature.
The platform economy adds a sharper turn, and it runs the other way from anything Habermas or Havel had to name. The Columbia legal scholar Tim Wu documented the architectural move through two centuries of media history, from the penny press of the 1830s through radio and television and into the platforms.9 In each iteration, free or cheap content captured human attention, and that attention was sold on to whoever wanted to act on it. The platform era completed the cycle by automating it. The exhaust of human conduct on a platform was discovered to predict behavior, the predictions could be sold, and a whole architecture grew whose product is not the service the user thinks they are using but a forecast of what the user will do, manufactured from the user’s own conduct and sold to whoever wants to act on it before the user does. Hold that against the inherited validation function and the inversion is exact. The citizen used to be the one who read what passed in the polity and vouched for it. In the attention architecture the citizen is the thing being read, the raw material the system harvests to predict, and the direction of the whole relationship has reversed. Validation once ran from the citizen outward onto the world. Now the world runs a model of the citizen and feeds back what the model predicts will hold attention, and the faculty of judgment the inherited architecture assumed as its input is being trained out of the input itself. The Greek former finance minister and political economist Yanis Varoufakis sharpens the inversion to the faculty itself: the platform is “a machine, a piece of capital, which we train to train us to train it to determine that which we want.”10
So the missing layer can be named, even if it cannot yet be built. A validation layer adequate to the new conditions would have to run at human resolution and human cadence, at the speed at which a citizen could actually read what passes in front of them and judge it. When the architecture that decides what a citizen sees is the same architecture that predicts and shapes what the citizen will want, the epistemic act and the sovereign act arrive through one channel and fail through one mechanism. You cannot vouch for the truth of what you can no longer independently see, and you cannot authorize as a free will a choice the channel has already steered.
Self-governing systems have always required a layer where some human or body of humans reads what the system is producing and vouches for it; the institutional forms change, the function does not. Aristotle, in Politics III.1, defined the citizen as whoever participates in deliberation or judgment, and locked the function to krisis. The Roman civis, the Enlightenment sovereign, the public Dewey diagnosed, the actor Arendt named, the reasoning public Habermas mapped, the human in Wiener’s loop: each is the same function under a different institutional form. The full lineage and citations are carried in Endnote 11.11 What has changed, for the first time in that whole span, is that the heaviest of the three components of coherent knowledge has been not merely overrun but architecturally reversed. Resolution and cadence can fail and the polity can muddle through; lose validation and the polity is making decisions without knowledge that can support them. The decisions still happen. They are no longer decisions in the sense the polity was built to make.
Where the function once ran from the citizen outward, it now runs from the system inward onto the citizen, and the layer that would turn it back is the layer that has not been built.
V.
Treat the three components as one architecture, because complexity science has been describing the architecture’s failure mode since 1963. Validation, grain, and clock must hold simultaneously for coherent knowledge to be possible, and the platform architecture has overrun all three at once. The mathematics says small failures in any one propagate through the whole.
I have watched one corner of this work for years in Consumer Packaged Goods. Many food manufacturers carry commodity risk that never touches an exchange, in the raw materials they buy to make what they sell, and over the last several years the engagements I have been part of have been working through a climate-driven shock to the global cocoa and coffee markets. The problem starts in the fields of West Africa, where much of the crop is harvested, and where the climate pressure sits on top of years of under-investment and missing modernization in the farming itself. One commodity, one region, and the disruption does not stay there; it cascades out to the rest of the world. The companies that come through this shock, mitigating risk or exploiting opportunity rather than being run over by it, are the ones whose readings hold on all three components at once: right scale, catching cadence, warranted enough to act on. Where one slips, the recommendation collapses in the predictable direction the missing component governs, and the manager finds themselves acting on a reading that no longer carries the conditions the decision was supposed to rest on. The shape has a name in the mathematics.
The American meteorologist Edward Lorenz gave the meeting point its mathematics. In a deterministic system governed by fixed nonlinear equations, a vanishingly small difference at the start can compound through the system’s interactions into a difference that swallows the answer at the end. Lorenz published the finding in 1963, and in 1972 he gave it its public name in a lecture to the American Association for the Advancement of Science, asking whether a butterfly’s wings in Brazil might set off a tornado in Texas.12 The image traveled and the mathematics did not. The mathematics says that in a system with strong nonlinear interactions and a fast-traveling signal, a perturbation far too small to see in one corner can produce a state change far too large to ignore at the other side, on a timeline shorter than any controller built for the system’s ordinary behavior was designed to answer.
The Lorenz claim does specific work for the architecture of coherent knowledge, and the work is structurally precise rather than merely illustrative. What Lorenz proved is that in a deterministic nonlinear system with strong feedback, a small undetected error in the input conditions propagates through the system into a state-prediction error far larger than the input error itself. A validation failure is, in those terms, an undetected error in the input conditions: the actor reads what is in front of them but has no warrant for trusting that what they read is what is actually there. The failure propagates through the knowledge stack (what the actor sees, what the actor trusts, what the actor decides) and arrives in the decision as something that no longer maps to the world. The popular reading of the butterfly fixes on input and output, on the tornado at the end of the chain. The mathematics points at the upstream component: an unwarranted input in a tightly-coupled system does not stay small, and a decision reached on the basis of that input is a guess the system has lost the ability to tell from a judgment. This is the condition the inherited sovereign now operates inside, and it shows up wherever the inherited cadence meets the new one: a regional virus becomes a planetary pandemic in nine weeks; a wholesale funding shock becomes a continental credit freeze inside a single day; a generated artifact becomes a political event in seventy-two hours. In each case the controller, calibrated to the cadence of legislative deliberation or board oversight, was structurally too slow to act before the next state change had already arrived, while the architecture that would have told it what it was looking at had already been overrun.
Take the pandemic of 2020. A pathogen that crossed from one host in one market became a presence on every continent, carried by the dense nonlinear interconnection of global air travel, before the institutions built to contain it had agreed on what they were looking at. The financial events of 2008 and again of 2023 are the same structure at a different scale; the first remains, in the most careful telling, politically unresolved.13 The information events of 2024 and 2025 are the same structure again. A single fabricated artifact is carried by the recommendation engines into a state change in the political weather before any editorial judgment can be brought to bear on whether it was true. The three components meet at the crossing, the warrant sits unbuilt at the load-bearing position, and a sovereign that cannot produce coherent knowledge at the intersection holds the name of sovereignty without holding the function.
VI.
An answer can be named in outline, even if its details are still unknown. The outline is an institution that produces coherent decisions, coherent knowledge in all three components, at the scales and the speeds the decision in front of it requires.
Such an institution would have to read at every resolution at which it is being asked to govern, from the single decision through the local condition through the national position to the planetary signal, without aggregating the structure out of the data before the reading begins. The inherited instruments each read cleanly at one scale and lose the problem at the others. An instrument that closes the gap reads at all of them at once, or it does not close the gap.
It would have to act at the cadence the environment actually moves at, which is the cadence of network propagation in one domain, of capital reallocation in another, of the climate horizon in a third. Not at a single speed, but at whatever speed the decision in front of it is moving, while holding legitimacy at every one of those speeds, which is the harder half, because speed without legitimacy is just a faster way to be wrong at scale.
Underneath both sits the heaviest move. The new instrument would have to hold a validation layer at human resolution and human speed inside an information architecture running at machine resolution and machine speed. Most of the architecture cannot be human-cadence, because reading at every scale at planetary reach and acting at network speed are tasks no human-cadence instrument can meet. The validation layer is the gate, the place inside the architecture where what the machines have read and proposed is held against human judgment before it becomes a decision the polity stands behind. The machines read and act at their speeds; humans vouch at theirs. That is the difference between governing and administering. Without that gate a new instrument can be fast and accurate and still not be a sovereignty, because it will be answering questions no human judgment ever validated. With it, the instrument could carry the legitimacy the inherited sovereign carries into conditions the inherited sovereign cannot reach. The structural need is real. So is the epistemic risk the same instrument inherits at the same hand.
One early instance of this architecture exists already, in commodity-risk decision systems for the food sector. A continuously-modeled global supply chain at the resolution of a climate-driven shock, for example, running at the cadence the market state-changes at, with a validation layer strong enough that a senior leader can act on the recommendation in the room. I have been inside that work for many years. The full instrument the polity needs is larger and not yet assembled. The pieces are real, as are the practitioners.
None of this replaces the instruments we have. Bounded threats still move at the speed of armies and trade. Firms still face decisions a board can weigh in a single meeting, and states still face problems a forum of states can work through across a season of negotiation. What has happened is not that the old conditions disappeared but that a new class of conditions arrived on top of them, asking a question the three were never tuned to answer. The American agrarian essayist Wendell Berry, writing about institutions that hollowed out from inside, returned to one finding that bears here. An institution is not recovered by reforming its administration. It is recovered when the practitioners take the work back up and do the thing the institution exists to do, at the scale where it can be done well. Whatever stands alongside the inherited sovereign will be built that way or not at all, by the people who do the actual work of reading and acting and validating, not by an administrative redesign of the bodies that have lost the resolution, the cadence, and the warrant.
Who those people are, what they owe the public realm, what the instrument looks like when it is running in the world, and whether it can be built at all, are real questions for another time. The moment is enough to hold on its own.
The state, the firm, and the multilateral body each answered the world they were built to answer. The world they now occupy is asking different questions, at a resolution and a cadence and a standard of warrant the old answers cannot reach. The instruments still work; the questions have moved. The layer that would vouch for what passes through them has been turned around to predict the one who used to do the vouching. What stands alongside the inherited sovereign will produce decisions a polity can defend, or the polity will continue to make decisions in name only.
Stephen DeAngelis
Princeton, NJ
June 2026
About the Author
Stephen F. DeAngelis is the founder, president, and CEO of Enterra Solutions and Massive Dynamics, two companies that apply artificial intelligence and advanced mathematics to complex enterprise challenges. His work spans international relations, national security, and commercial technology, with visiting research affiliations at Princeton University, Department of Chemistry, the Computing and Computational Sciences and National Security Directorates of the Oak Ridge National Laboratory, the Software Engineering Institute at Carnegie Mellon University, and the MIT Computer Science and Artificial Intelligence Laboratory. He holds patents in autonomous decision science.
1. Zachary D. Carter, The Price of Peace: Money, Democracy, and the Life of John Maynard Keynes. New York, Random House, 2020. Carter’s book is a recovery of Keynes the political economist against the Keynesian-mechanism caricature, reading the Bretton Woods summer through the man rather than through the model that later took his name. https://www.penguinrandomhouse.com/books/563378/the-price-of-peace-by-zachary-d-carter/
2. The most-documented single case is the Pope Francis white-puffer-coat AI image (Midjourney, March 24-25, 2023). The single most-cited tweet carrying the image (@skyferrori, March 25, 2023) reached 28.1 million views and approximately 197,000 likes by mid-April 2023, with the image spreading across Reddit, Twitter, Facebook, and Instagram within hours of initial posting (Romano, AI Incident Database report 3606, January 2024; corroborated by New York Times coverage April 8, 2023, and Boston Globe March 29, 2023). The Pentagon explosion deepfake (May 22, 2023) briefly moved U.S. equity markets, with the S&P 500 dropping approximately 0.3% in the minutes following the image’s circulation before the falsehood was identified. The synthetic Biden-voice robocall in the New Hampshire Democratic primary (January 21-22, 2024) reached an estimated 5,000 to 25,000 New Hampshire residents per the New Hampshire Attorney General’s investigation. Each of the three cases documents platform-distribution reach measured in the millions to tens of millions of impressions within twenty-four hours of generation.
3. Nils Gilman, “Governing in the Planetary Age,” Noema Magazine, March 9, 2021. Gilman argues that the scale of planetary challenges is incommensurate with the governing capacity of the nation-state, which is “too small for the big problems of life and too big for the small problems,” a structural rather than volitional failure of fit. https://www.noemamag.com/governing-in-the-planetary-age
4. John Dewey, The Public and Its Problems. New York, Henry Holt and Company, 1927, Chapter 4 (“The Eclipse of the Public”). Dewey, responding to Walter Lippmann, accepts that the public “seems to be lost; it is certainly bewildered” and locates the failure in the indirectness and complexity of modern consequences rather than in any incapacity of citizens.
5. The pandemic time scales draw on the 2020 COVID-19 propagation record (confirmed cases reported across more than a hundred countries between January and March 2020 per WHO Situation Reports 1 through 60). The legislative-cadence figures draw on Congressional Research Service analysis of reconciliation legislation enacted since 1980, which documents intervals from initial budget-resolution adoption to enactment ranging from 27 days (Omnibus Budget Reconciliation Act of 1990) to 384 days (Tax Increase Prevention and Reconciliation Act of 2005), with a mean of approximately 150 days (CRS Report RL30458, The Budget Reconciliation Process: Timing of Legislative Action). Hughes and Carlson’s 2015 American Politics Research analysis of bills 1949-2010 documents median time-to-enactment ranging from 63 to over 200 days across Congresses, with most substantive measures requiring six to eighteen months. The point in the body is structural rather than precise: the asymmetry between information cadence, threat cadence, and legislative cadence runs to many orders of magnitude in every well-documented planetary crisis of the last decade.
6. Hannah Arendt, The Human Condition. Chicago, University of Chicago Press, 1958, Section 6 (“The Rise of the Social”). The public-freedom passages that hold up against that conquest are in On Revolution. New York, Viking, 1963. Arendt’s distinction between political action and administration is the distinction the cadence mismatch in section III converts from a philosophical claim into a structural one. https://press.uchicago.edu/ucp/books/book/chicago/H/bo29137972.html
7. The evidentiary standard the body is asserting is bounded as follows: no published evaluation across the three relevant literatures, namely primary (foundation-model laboratory disclosures), academic (peer-reviewed interpretability and evaluation literature), and regulatory (statutory and standards-body risk frameworks), has documented a general-purpose foundation model that meets the warrant condition the body claims is missing, namely a documented capacity for the model itself to vouch reliably for the truth of what it produces. Convergent findings on this gap include: Anthropic’s published model cards on Claude releases (discussions of hallucination and faithfulness limits, March 2023 forward); OpenAI’s system cards on GPT-4 (March 2023), GPT-4o (May 2024), and successor models (refusal failures and confabulation documented in each release); Stanford HAI’s AI Index annual reports (2023, 2024, 2025) on benchmark performance and known failure modes; Bender, Gebru, McMillan-Major, and Shmitchell, “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?”, FAccT ’21, Association for Computing Machinery; the EU AI Act (Regulation 2024/1689, entered into force August 1, 2024) and its risk-tiering of opaque general-purpose models; the U.S. Executive Order 14110 on Safe, Secure, and Trustworthy Artificial Intelligence (October 30, 2023) and the NIST AI Risk Management Framework 1.0 (released January 26, 2023); and the UK AI Safety Institute’s evaluation work since November 2023. The convergence is across primary, academic, and regulatory sources; the body is not asserting unanimous consensus, only that no source across the three literatures contradicts the gap. A single peer-reviewed or regulator-documented model meeting the warrant condition would falsify the claim.
8. Jürgen Habermas, Strukturwandel der Öffentlichkeit (1962); English translation by Thomas Burger with Frederick Lawrence, The Structural Transformation of the Public Sphere. Cambridge, MIT Press, 1989, at 206. https://mitpress.mit.edu/9780262581080/the-structural-transformation-of-the-public-sphere/
9. Tim Wu, The Attention Merchants: The Epic Scramble to Get Inside Our Heads. New York, Knopf, 2016. Wu traces the structural logic of attention-as-commodity across two centuries of media, from Benjamin Day’s New York Sun in 1833 through the rise of platform advertising, and reads the platform era as the latest iteration of a recurring extraction cycle rather than a digital novelty. https://www.penguinrandomhouse.com/books/234876/the-attention-merchants-by-tim-wu/
10. Yanis Varoufakis, Technofeudalism: What Killed Capitalism. London, The Bodley Head, 2023. https://www.penguin.co.uk/books/451795/technofeudalism-by-varoufakis-yanis/9781529926095
11. The body compresses a twenty-four-century lineage into one sentence. What follows is the apparatus the compression rests on. The lineage names the history of how each epoch understood the validation function (the heaviest component of coherent knowledge) and the institutions through which the polity exercised it.
Classical foundations. Aristotle defines the citizen at Politics III.1, 1275b16 to 20: “whoever is entitled to participate in an office involving deliberation or judgment is a citizen of that polis.” The definition is functional, not territorial. The citizen’s constitutive attribute is krisis, judgment exercised in the deliberative and judicial assemblies. He further anticipates the distributed-validation argument at 1281b, noting that the assembled multitude may collectively judge better than the wise individual, “as a feast to which all the guests contribute is better than a banquet furnished by a single man.” The Roman civis equipped with provocatio ad populum institutionalizes the architecture at the procedural level.
Enlightenment social contract. Rousseau radicalizes the Aristotelian function in the Social Contract (1762). Members of the body politic are “called citizens, as sharing in the sovereign power, and subjects, as being under the laws of the State”; the same person is both, and the citizenship is the moment of judgment. Sovereignty is defined as “nothing less than the exercise of the general will” and is by construction inalienable. Jefferson grounds the abstract argument in institutional architecture (letters to Edward Carrington, 16 January 1787; to Charles Yancey, 6 January 1816). Madison engineers the scaling response: in Federalist No. 10 and No. 51, the structural design of the extended republic becomes the mechanism by which aggregated citizen judgment performs the validation function at a scale beyond which the Athenian assembly form would hold.
Twentieth-century diagnoses. Three figures named distinct failure modes of citizen validation under industrial-and-mass-media conditions. Dewey in The Public and Its Problems (1927) diagnoses the self-recognition failure: when consequences become too indirect to trace, the public “seems to be lost; it is certainly bewildered.” Arendt in The Human Condition (1958) names the collapse of politics into administration; the citizen as actor in the public realm is replaced by the subject as job-holder, and “society has thus invaded and conquered the public realm.” Habermas in Strukturwandel der Öffentlichkeit (1962; English translation 1989) identifies the institutional mechanism: the bourgeois public sphere undergoes structural transformation through the interpenetration of state and economy; at 206, “publicity loses its critical function in favor of a staged display.”
Cybernetic translation. Wiener in Cybernetics (1948) and The Human Use of Human Beings (1950) translates the validation argument into feedback-controlled systems. The human in the loop is the element whose judgment defines correct performance; remove the human and the system has no reference against which to validate its output. Licklider in “Man-Computer Symbiosis” (1960) sharpens the architectural claim: “In the anticipated symbiotic partnership, men will set the goals, formulate the hypotheses, determine the criteria, and perform the evaluations. Computing machines will do the routinizable work.” Engelbart in “Augmenting Human Intellect” (1962) establishes the augmentation-versus-automation principle.
Contemporary inversion. The continuity breaks where the attention architecture reverses the direction of validation. Wu (2016) documents the conversion in industry-historical terms across two centuries of media: citizen behavior becomes the raw material from which prediction and engagement products are manufactured. Varoufakis (2023) sharpens the diagnosis to the faculty itself; the politēs who exercises krisis in the assembly becomes the cloud serf who asks the platform what to want. The inversion documented here is the first historical moment at which the heaviest component of coherent knowledge has been not merely overrun but architecturally reversed.
12. Edward N. Lorenz, “Deterministic Nonperiodic Flow,” Journal of the Atmospheric Sciences 20, no. 2 (March 1963), 130 to 141; and the 1972 address to the American Association for the Advancement of Science, “Predictability: Does the Flap of a Butterfly’s Wings in Brazil Set Off a Tornado in Texas?” https://journals.ametsoc.org/view/journals/atsc/20/2/1520-0469_1963_020_0130_dnf_2_0_co_2.xml
13. Adam Tooze, Crashed: How a Decade of Financial Crises Changed the World. New York, Viking, 2018. Tooze treats the 2008 crisis and its aftermath as politically unresolved and ongoing rather than a closed chapter. https://www.penguinrandomhouse.com/books/301357/crashed-by-adam-tooze/
Share this post
