Belief as Infrastructure, the full paper · Future Proof Intelligence

Belief as Infrastructure

How conviction becomes coordination, coordination becomes a market, and why, in an era of abundant machine output, what humans believe and whom they believe becomes the binding constraint.

Future Proof Intelligence. Research. No. V. MMXXVI

Abstract

Belief is usually treated as a soft factor: a matter of culture, sentiment, or opinion that sits on top of the hard machinery of an economy and occasionally disturbs it. This paper argues the opposite. Belief is a load bearing layer. Money, property, law, authority, standards, brands, ratings, and networks are not things that belief decorates. They are things that belief constitutes. Each is a structure of collective recognition: real because it is believed, and real only for as long as it is believed by enough people who believe that others believe it too.

We trace a single mechanism across these structures. Private conviction becomes common knowledge, common knowledge becomes a self enforcing convention, and a convention that enough people are sure others will keep honouring becomes a market. The same mechanism explains how a currency holds and how it collapses, why a standard that is believed to be the standard becomes the standard, and why trust behaves like capital that is slow to accumulate and fast to destroy.

We then turn to the present. When machine output becomes abundant and nearly free, production stops being the scarce input. The scarce, load bearing input becomes credence: whether a signal can be believed, and whom. The economy is quietly repricing around verified belief. Institutions that understand belief as infrastructure, rather than as a soft factor, will be the ones still standing when the repricing completes.

1. The Category Error

1.1 What we mean by belief

There is a habit, deep in the way modern institutions describe themselves, of sorting the world into the hard and the soft. The hard is the part that can be measured: capital, output, throughput, the balance sheet, the supply chain, the model weights. The soft is everything that resists a number: morale, reputation, narrative, confidence, trust, conviction. The hard is treated as the structure. The soft is treated as the weather that passes over the structure and is sometimes inconvenient.

This paper is about a category error inside that habit. The thing filed under soft is, in many of the cases that matter most, the structure itself. Not a force acting on the system. The material the system is made of.

We need to be precise about the word, because belief carries connotations the argument does not need. We do not mean faith in the religious sense. We do not mean private opinion, the thing a person holds in the quiet of their own head and could change tomorrow at no cost to anyone. We mean something narrower and harder: credence that coordinates action. A belief, in the sense that does economic and institutional work, is an expectation about the world that a person acts on, and that becomes consequential precisely because other people are acting on the same expectation, and because each of them is, to some degree, aware that the others are.

It is worth holding three distinctions clearly, because the rest of the paper depends on them. The first is between a belief that is held and a belief that is held in common. A thousand people who each privately suspect the same thing, with no awareness of each other, form no structure at all; the same thousand, each aware that the others suspect it, form something that can move. The second is between belief about the world and belief about other people's beliefs about the world. The economically load bearing kind is almost always the second, the recursive kind, and the recursion is not a complication to be abstracted away. It is the mechanism. The third is between conviction and consequence. This paper is not concerned with whether a belief is true, noble, or sincerely felt. It is concerned only with whether enough people act on it, in mutual awareness, that it acquires the properties of a structure. A false belief, commonly held and acted on, is more load bearing than a true one held privately by a few. That is uncomfortable and it is the point.

That last clause is the whole engine. A belief held alone is a private state. A belief that many hold, and that each holds partly because they expect the others to hold it, is no longer a private state. It is a structure. It has load bearing properties. You can build on it, lend against it, organise around it, and you can also watch it fail in ways that have nothing to do with anyone changing their mind in isolation.

1.2 The proof that is in everyone's pocket

The cleanest demonstration is the one most people carry every day and never look at directly.

Consider a banknote, or, more honestly for 2026, a number in a banking application that represents a balance. Ask what makes it worth what it is worth. The paper has almost no intrinsic value. The number has none at all; it is a record in a database. The note is not redeemable for gold or for anything else with independent worth. Its value does not come from its substance. It comes from the fact that the person you hand it to will accept it, and they will accept it because they expect the next person to accept it, and so on, in a chain that has no anchor at the bottom except the expectation itself.

The philosopher John Searle gave this its sharpest formulation. In The Construction of Social Reality and later in Making the Social World, he argued that an entire class of facts, the institutional facts, exist through a single mechanism: a status function, captured in the formula that some X counts as Y in a context C. A piece of paper counts as a twenty pound note. A person counts as a head of state. A line on a map counts as a border. These status functions carry what Searle calls deontic powers: rights, duties, entitlements, obligations. And they are held in place by collective intentionality, the shared recognition of a population that treats the X as the Y. The note is money because we collectively count it as money. It is not money in the way that water is wet. It is money in the way that a king is a king: by recognition, and for exactly as long as the recognition holds.

This is not a philosophical curiosity. It is the load bearing observation of this paper. The most concrete, quantitative, hard edged object in a modern economy, money, is at its foundation a crystallised belief. Everything denominated in money inherits that foundation. The hardest number in the system rests on the softest thing in the catalogue. The category error is to think that makes money fragile and embarrassing. It does not. It makes belief structural.

1.3 Crystallised belief

We will use a phrase throughout this paper and it is worth defining it once, carefully. By crystallised belief we mean a belief that has been given a durable external form, such that it now coordinates the behaviour of people who never deliberated about it and may never examine it. A banknote is crystallised belief. So is a property deed, a credit rating, an industry standard, a constitution, a brand, a currency peg, a passport, and a price.

The crystallisation matters. A belief in someone's head is fluid and private. The same belief, written into an instrument that thousands of people treat as authoritative, becomes something else. It becomes infrastructure: a thing other things are built on top of, that most participants take as given, that has interfaces and dependencies and failure modes, and that, when it fails, brings down what was built on it.

The rest of this paper is, in one sense, a single argument applied at increasing scale. We start with money because it is the proof everyone already half knows. We then show that the same mechanism builds institutions, law, authority, standards, and markets. We examine how the crystallisation happens, the phase transitions by which private conviction becomes a public structure. We look at how this infrastructure fails, because its failure modes are unlike those of physical infrastructure and are routinely misread. And then we turn to the present, where a specific technological shift is making the belief layer the scarce and binding constraint on economic life for the first time in a way that is visible to everyone at once.

2. Money as the Canonical Structure

2.1 Credit means I believe

The economist Geoffrey Ingham, in The Nature of Money, made an argument that is still under absorbed outside economic sociology. Money, he contends, is not a commodity that markets happened to settle on for convenience. It is a social relation: specifically, a structure of credit and debt, a politically constituted promise to pay, expressed in an abstract unit of account. The unit of account, the pound or the euro or the dollar as a measure rather than as a coin, is the deep layer. It is, in Ingham's phrase, a collective representation: a shared abstraction the whole society agrees to reckon in.

The etymology is not decoration. Credit comes from the Latin credo: I believe. A creditor is one who believes a debtor will pay. A monetary system is a dense web of such beliefs, formalised, transferable, and enforced. When you accept money for goods, you are extending credit to the entire system. You give up something real for a token, on the belief that the token will be honoured by strangers you will never meet. The system works because that belief is general, and it is general partly because it is general: each person believes because they observe everyone else behaving as if they believe.

This is the first appearance of a pattern we will see repeatedly. Belief in these structures is not merely shared. It is reflexive. Each participant's belief is, in part, a belief about other participants' beliefs. That recursion is what gives the structure its strength and also what gives it its characteristic mode of collapse.

There is a familiar objection at this point, and disposing of it cleanly is worth a paragraph because the objection recurs in every domain this paper touches. The objection is that money used to be backed by gold, and gold has intrinsic value, so the belief account is at best a description of a recent and possibly unsound arrangement. The objection does not survive contact with the history. A gold backed currency did not remove the belief layer. It relocated it. A note redeemable for gold is worth its face value only if the holder believes the issuer will honour the redemption, and that everyone else believes it too, so that the note circulates rather than being presented all at once. The classical gold standard did not run on metal. It ran on a shared and reflexive confidence that convertibility would be maintained, and it broke, repeatedly and across many countries, at the precise moments that confidence broke, not at the moments the metal physically moved. Gold itself, examined honestly, is the same structure one layer down: a substance whose monetary value across millennia rests overwhelmingly on the durable, near universal expectation that others will continue to treat it as valuable. There is no terminating layer of intrinsic worth at the bottom of money. There is only belief, more or less durable, more or less reflexive, more or less crystallised. The history of money is not a fall from a hard standard into a soft one. It is the same material throughout, becoming progressively more honest about what it is made of.

2.2 The bank run as a belief phase transition

Nothing illustrates the load bearing nature of belief like the moment it withdraws.

A bank run is usually described in the language of liquidity: a bank lends long and borrows short, cannot meet simultaneous withdrawals, and fails. That is mechanically true and explanatorily incomplete. The deeper description is that a bank is solvent precisely as long as its depositors believe it is solvent and therefore do not all ask for their money at once. The belief is not a commentary on the bank's health. For a fractional reserve institution, the belief is a component of the bank's health. Depositors who individually conclude the bank may fail, and act on it, cause the failure they predicted. The prediction is self fulfilling because the thing being predicted is itself partly made of the predictions.

Diamond and Dybvig modelled this formally in 1983 and won a share of the Nobel Memorial Prize for it in 2022: a bank has two equilibria, one in which everyone expects everyone else to stay and the institution is stable, and one in which everyone expects everyone else to run and the institution is destroyed, and nothing about the bank's underlying loans needs to change to move the system from the first to the second. The state that obtains is selected by belief about belief. Deposit insurance does not work primarily by paying people out. It works by changing what people believe others will do, so that the run equilibrium becomes unreachable. The insurance is, in the deepest sense, a belief technology.

Currency crises and hyperinflations are the same phenomenon at the scale of a whole monetary system. A hyperinflation is not usually, at its core, a story about the printing press alone. It is a story about a population ceasing to believe that the unit of account will hold its meaning, and acting on that disbelief by spending the currency the instant they receive it, which destroys the very stability whose absence they were anticipating. The money supply is the mechanism. The collapse of the collective representation is the event. A currency dies the way a language dies: not when the last note is printed, but when people stop reckoning in it.

2.3 Why this generalises

It would be easy to read the previous section as a set of observations about finance. It is not. Money is simply the case where the mechanism is most exposed, because money has no physical pretence to hide behind. A government bond, a corporate equity, a derivative, a currency peg, an insurance pool, a clearing house: each is a more elaborate instance of the same thing, an instrument whose value is the collectively held and reflexively reinforced expectation that it will be honoured. The 2008 financial crisis was, told at the right altitude, a story about a class of instruments that were believed to be safe, a belief that was crystallised into ratings and capital rules and risk models, and the very rapid, very nonlinear withdrawal of that belief once a few participants stopped believing and others noticed them not believing.

The point of starting with money is that it forecloses an objection. Nobody can claim that money is a soft factor sitting on top of the real economy. Money is the substrate of the real economy, and money is crystallised belief. Once that is conceded, the rest of the argument is a matter of showing how far the same substrate extends.

3. The Belief Stack

3.1 Institutions are durable shared mental models

Douglass North, in Institutions, Institutional Change and Economic Performance, defined institutions as the rules of the game in a society: the humanly devised constraints, both formal rules and informal norms, plus their enforcement, that structure economic and political interaction. Institutions, in his account, are the incentive structure of an economy, and the long run performance of an economy is largely the story of which institutions it has and how they change.

What is sometimes lost in summaries of North is the role of belief inside the definition. The informal constraints, the customs, codes, norms, and conventions, are not written anywhere. They exist as shared expectations: a population's common understanding of what is done and what is not, and what will happen to you if you do what is not done. North was explicit that participants act through mental models, the theories and ideologies and frames they use to interpret the game. An institution, at its core, is a shared mental model that has become durable enough, and common enough, to coordinate strangers. It is crystallised belief with an enforcement mechanism bolted on. The enforcement matters, but the enforcement is itself only credible because it is believed to be credible. A law that no one expects to be enforced does not constrain behaviour, regardless of what is written.

3.2 Authority is belief in legitimacy

Max Weber, in Economy and Society, gave the canonical analysis of why people obey. He distinguished power, the ability to compel, from authority, or legitimate domination, the situation in which people comply because they believe the command is rightful. He identified three pure bases for that belief: tradition (it has always been so), charisma (this person is extraordinary), and legal rational grounds (the rule was made by a procedure we accept). The common element across all three is the word belief. Authority is the condition in which compliance is given because legitimacy is believed in, not because force is continuously applied.

This is not a normative claim about good government. It is an engineering claim about cost and durability. Power that rests on belief in its legitimacy is extraordinarily cheap to run, because most people comply most of the time without being made to. Power that rests only on the threat of force is extraordinarily expensive, because every act of compliance has to be purchased with the credible prospect of coercion, and the moment the coercion looks less credible the structure starts to dissolve. Regimes do not usually fall because the army is defeated. They fall when enough people simultaneously stop believing the regime will last, and act accordingly, and observe others doing the same. The belief was the load bearing element. The visible collapse is downstream.

3.3 Property, contract, and the rest of the stack

Once the pattern is visible, it appears everywhere in the architecture of an economy.

A property right is not a physical fact about a parcel of land. It is a collectively recognised and enforced status function, in exactly Searle's sense: this person counts as the owner, and the belief that this is so, backed by a registry and a court and a population that mostly honours it, is what lets the owner borrow against it, sell it, or build on it. A contract is a promise made enforceable by a shared belief that the system will compel performance. A brand is a compressed belief about quality and identity that lets a buyer skip an investigation they cannot afford to run every time. A credit rating is an outsourced belief: a population of investors agreeing to treat one organisation's opinion as a coordination point so they do not each have to form their own. A central bank's inflation target works to the precise extent that it is believed; the credibility is the policy instrument, and the technical machinery exists largely to make the belief sustainable.

We can now state the structure plainly.

Figure 1. The belief stack. At the base sits private credence: what an individual expects. Above it sits common knowledge: shared credence that each participant knows the others share. Above that sits convention: a self enforcing pattern of behaviour that holds because each participant expects the others to keep honouring it. Above that sits crystallised belief: conventions given durable external form as money, law, property, standards, ratings, and brands. At the top sits the market: priced, traded, scaled coordination resting on every layer beneath it. Each layer is built on the one below. Remove enough of any lower layer and everything above it descends without anything in the upper layers having changed.

The reason the hard and soft distinction is a category error is now precise. The "hard" objects of the economy are the upper layers of this stack. The "soft" factor of belief is not sitting beside them or on top of them. It is every layer, all the way down. The hardness of the upper layers is borrowed entirely from the stability of the belief beneath them.

3.4 The deepest layer: measurement, language, and time

It is tempting to think the stack bottoms out somewhere in something physical. It does not, and the cases where it most obviously does not are the ones we lean on hardest without noticing.

Consider measurement. A market cannot price anything without a unit, and a unit is a crystallised belief of the purest kind: a collective agreement to treat a particular quantity as the reference, sustained because everyone expects everyone else to keep treating it as the reference. The metre, the kilogram, the second, the accounting period, the index, the benchmark rate: each is a focal point that became a standard because enough participants believed enough others would use it. The history of metrology is, read at the level of this paper, a history of belief being negotiated into infrastructure, which is why redefining a base unit is a diplomatic and institutional act and not merely a scientific one. Every quantitative claim in economics is denominated in units that are themselves crystallised conventions. The number does not escape the belief layer. It is expressed in it.

Language is the same case at greater depth. The meaning of a word is sustained by nothing more than the durable, reflexive expectation that others will use it the same way. A contract is enforceable only because the parties, the court, and the population share a convention about what its terms mean, and that convention is a structure of common knowledge with exactly the properties this paper has been describing. When the shared meaning of a key term erodes, the contract does not become ambiguous gradually and gracefully. It becomes contestable suddenly, the moment the divergence in interpretation becomes common knowledge, which is the same nonlinear signature we will return to in section 5.

Time itself, as economies use it, belongs here. A fiscal year, a settlement date, a maturity, a deadline: these are not features of the physical world. They are collectively honoured conventions that let strangers coordinate across duration, and they hold for the same reason every other layer holds. The point of pressing the stack this far down is not to be clever about foundations. It is to remove the last place a reader might hope to stand outside the argument. There is no floor of brute fact beneath the economy on which belief merely rests. Measurement, meaning, and time, the apparatus with which we describe the economy at all, are themselves crystallised belief. The argument does not have an outside.

4. How Conviction Becomes a Market

4.1 The three transitions

The interesting question is not whether belief is structural. Section 3 settles that. The interesting question is mechanical: by what process does a private conviction, the cheapest and most fluid thing in the world, turn into a market, the most concrete? There are three transitions, and each is a phase change with a threshold.

Transition one: private credence to common knowledge. A belief held by many people in isolation has almost no coordinating power. The same belief becomes powerful at the moment it becomes common knowledge, in the technical sense the philosopher David Lewis gave the term in Convention: not merely that everyone believes X, but that everyone believes that everyone believes X, and so on. The difference is enormous. Many people privately doubting an institution change nothing as long as each believes the others still trust it. The same people change everything the instant the doubt becomes mutually visible. This is why the publication of a fact, the moment it goes from privately suspected to commonly known, can move a market that the underlying fact, privately held, did not move at all. The fact did not change. Its epistemic status did. Common knowledge is a different physical state of a population than distributed private knowledge, and structures respond to the state, not to the aggregate.

Transition two: common knowledge to convention. Thomas Schelling, in The Strategy of Conflict, showed that people coordinate without communicating by converging on focal points: the salient option, the one each expects the others to expect. His example endures because it is so ordinary. Strangers told to meet in a city, with no way to communicate, disproportionately succeed, because there is an obvious place and an obvious time and each reasons that the others will reason the same way. A convention is a focal point that has stabilised: a pattern everyone follows because everyone expects everyone to follow it, self enforcing because deviating is costly when everyone else is coordinated. Conventions need no central enforcer once established. They are held up entirely by the recursive expectation. This is the cheapest coordination technology that exists, and almost all of economic life runs on it.

Transition three: convention to market. A market is a convention that has been priced. Once a convention is stable enough that participants will commit resources on the strength of it, the convention can carry contracts, then instruments, then prices, then derivatives of those prices. The market is the convention monetised. And because, per section 2, money is itself crystallised belief, a market is belief priced in a unit that is itself belief. It is belief all the way down, and that is not a weakness in the description. It is the most accurate description there is.

To make the three transitions concrete, follow one belief through all of them. A small number of people come to think that a particular kind of claim, say a verifiable record of where a piece of work came from, is going to matter. At first this is private credence: scattered individuals, each with a hunch, coordinating nothing. The transition begins when the hunch becomes visible and mutual, through writing, through standards meetings, through a few institutions publicly committing, so that the people who suspected it now know the others suspect it too. That is transition one, and nothing material has been built yet; what has changed is the epistemic state of a population. Transition two occurs when behaviour starts to converge: organisations begin to act as if the record will be expected of them, not because anyone compels it but because each now expects the others to expect it, and the salient way of doing it becomes the way it is done. A convention has formed, self enforcing, with no central enforcer. Transition three occurs when participants will commit real resources on the strength of the convention: when contracts reference it, when it is priced into procurement, when instruments are written that pay out depending on it. At that point a market exists, and it exists on top of a convention, which exists on top of common knowledge, which exists on top of what was, a short time earlier, a private hunch held by a few. No step in that sequence required the underlying claim to be true. Every step required it to be believed, in mutual awareness, by enough people. This is not a hypothetical. It is the abstract shape of how essentially every new coordination layer in economic history has come into being, and, as section 6 will show, it is the shape of the contest happening right now over what a believable signal is.

4.2 The standard as the purest case

There is one structure where the entire mechanism is visible in a single move, and it is the most important structure for the argument that follows: the standard.

A standard, technical or institutional, has a property that ordinary goods do not. Its value to any participant rises with the number of other participants who adopt it, and its adoption by any participant is driven mostly by the expectation of adoption by others. This is the network effect, but stated at the level of belief it is sharper: the standard that is believed to be the standard becomes the standard. The expectation is not a forecast of an independent outcome. The expectation is the mechanism that produces the outcome. A format, a protocol, a certification, a measurement, a benchmark, a settlement system: each becomes load bearing not when it is proven best but when enough participants believe enough other participants will treat it as the reference. After that point it is extremely hard to dislodge, not because it is technically superior, often it is not, but because dislodging it requires re coordinating a belief that has crystallised into everyone's expectation of everyone else.

This is the reason the contest to set a standard is so much more consequential than the contest to win a market inside an existing standard. Winning inside a standard is a contest over a flow. Setting the standard is a contest over the riverbed. Whoever shapes the convention that everyone else's expectations settle onto has built a piece of infrastructure that the rest of the economy then has to route through, and they have built it out of nothing more material than other people's beliefs about each other.

4.3 Reflexivity: belief is an input to reality, not a mirror of it

Two more thinkers complete the mechanism, and they matter especially for the era this paper is heading towards.

John Maynard Keynes, in Chapter 12 of The General Theory, described professional investment with an analogy that has outlived almost everything else in the book. He compared it to a newspaper competition in which readers must pick the faces a panel will judge most beautiful, so that the rational entrant does not pick the faces they find beautiful, nor even the faces they think others find beautiful, but devotes their intelligence to anticipating what average opinion expects average opinion to be. Value, in a market dominated by this dynamic, is not an estimate of a fundamental. It is a recursive estimate of other people's estimates. The convention holds because participants assume it will hold, and inquire no further, until something makes them inquire, at which point the convention can evaporate at a speed that the underlying fundamentals never moved at.

George Soros, who traded on this for a career, named the deeper version reflexivity. In his account, the belief participants hold about a market is not a passive reflection of that market's reality. It is an input to that reality. Optimistic belief about a company can raise its share price, which lowers its cost of capital, which improves its fundamentals, which appears to vindicate the optimism, which raises the belief further. Pessimism runs the same loop in reverse. The map redraws the territory it claims to depict. This is the property that makes belief infrastructure different in kind from physical infrastructure. A bridge does not get stronger because more people believe it is strong. A currency, a bank, a standard, a platform, and a reputation do exactly that, and weaker the same way, because in their case the belief is not a description of the load bearing element. The belief is the load bearing element.

5. The Physics of Trust

5.1 Trust is measurable capital, not sentiment

If belief is the substrate, trust is the form of it that economies have studied most directly, because its absence is so expensive that it can be priced.

Kenneth Arrow, in The Limits of Organization, observed that virtually every commercial transaction contains within itself an element of trust, certainly any transaction conducted over a period of time, and that much of the economic backwardness in the world can be explained by the lack of mutual confidence. He called trust a lubricant of a social system: it is the thing that lets exchange happen without every promise being fully specified, monitored, and enforced. Where trust is present, transaction costs collapse. Where it is absent, they explode, and a large fraction of the activity that trust would have permitted simply does not occur, because the cost of doing it safely exceeds the gain.

This is not a metaphor and it is not unquantified. Francis Fukuyama, in Trust, argued that the capacity of a society to form large scale economic organisations without kinship or coercion is a direct function of its level of generalised trust, and that low trust societies pay a structural tax in monitoring, contracting, and litigation that high trust societies do not. Zak and Knack, in a study published in The Economic Journal in 2001, found that measured generalised trust is positively associated with investment and economic growth across countries. Trust behaves like a stock of capital: it lowers the cost of every future transaction, it can be invested in, it can be drawn down, and a society or an organisation that depletes it pays for the depletion in slower, more expensive, more defensive economic life. The reason this is rarely on a balance sheet is not that it is immaterial. It is that it has historically been hard to denominate. Difficulty of measurement is not evidence of softness. It is a gap in the instruments.

5.2 The thin film: why belief infrastructure fails nonlinearly

Belief infrastructure has a failure signature that physical infrastructure does not, and misreading it is one of the most expensive errors institutions make.

Physical infrastructure degrades roughly in proportion to stress. A bridge under one and a half times its rated load is more strained than under its rated load, in a way that is continuous and, with inspection, legible. Belief infrastructure does not behave like this. It is closer to a thin film under tension. It holds, holds, holds, exhibiting very little visible change across a wide range of stress, and then it does not hold, and the transition between those two states is fast, discontinuous, and in the worst cases close to total. The bank in section 2 is solvent and solvent and solvent and then, within hours, gone, with no intermediate state in which it is partly run. The currency reckons and reckons and reckons and then no longer reckons. The standard is unquestioned and unquestioned and then abandoned.

The reason is the recursion. Because each participant's belief depends on their reading of other participants' beliefs, the system has the structure that produces tipping points. Below a threshold of doubt, the doubters each assume the others still trust, and so they continue to behave as if they trust, and the structure shows no sign of strain. Above the threshold, the doubt becomes mutually visible, each doubter now expects the others to act on it, and the rational response to that expectation is to act first, which moves the whole population across the boundary almost at once. There is very little useful warning in the observable behaviour of the system, because the system's stability and its fragility look identical right up until the transition. An institution that reads "no visible problems" as "no problem" has misunderstood the physics of the thing it is standing on.

5.3 The asymmetry

Trust, and belief infrastructure generally, has a second property that compounds the first. It is profoundly asymmetric in time. It accumulates slowly, through long sequences of consistent behaviour in which a counterparty does what it said it would do, especially when it would have been cheaper not to. It is destroyed quickly, sometimes in a single observed instance of a structure not honouring what people believed it would honour. Years of credibility can be undone by one event that becomes common knowledge.

The asymmetry is not unfair. It is informationally correct. A long record of honoured promises is consistent with a counterparty that is sometimes lucky; a single dishonoured promise, made common knowledge, is strong evidence about the counterparty's true constraints, and the rational population updates hard and fast. This is why reputational damage is so disproportionate to the act that caused it, and why the most valuable thing a long lived institution owns is almost never on its balance sheet. It is the accumulated belief that it will behave the way it has always behaved, an asset that took decades to build, that can be spent in an afternoon, and that, once spent, is far more expensive to rebuild than it ever was to maintain.

5.4 The institutions that exist only to manufacture belief

There is a class of institution whose entire function is to produce and maintain belief on behalf of others, and noticing that this class exists at all is one of the strongest confirmations of the thesis. An auditor produces nothing a buyer consumes; an auditor produces a believable signal that a set of accounts can be relied upon, and the value of the firm is precisely the credibility of that signal. A rating agency manufactures a coordination point so that millions of investors do not each have to form an independent view, and the agency's franchise is worth exactly what the market's belief in its ratings is worth, which is why a rating agency that is caught having signed off on what later fails suffers damage out of all proportion to the fees involved. A central bank, stripped to its function, is an institution for the production and defence of belief in a unit of account. A standards body manufactures the focal point that lets incompatible systems interoperate. A court manufactures the credible expectation that promises will be enforced, which is what makes promises worth making.

These institutions are not peripheral services bolted onto a real economy that could function without them. They are load bearing precisely because the economy is made of belief and belief at scale requires dedicated machinery to manufacture, certify, and defend. The existence of an entire professional civilisation devoted to producing trustworthy signal, accountants, auditors, raters, registrars, notaries, regulators, certifiers, is not evidence that trust is a soft afterthought. It is evidence that trust is so structurally essential that societies build expensive permanent institutions whose only product is the belief that something can be relied upon. Hold that observation. It is the one that the closing sections of this paper will need, because the question the AI era forces is not whether such institutions matter. It is which of them, and which new ones, the world will believe when the old signals stop carrying.

We now have the complete machine. Belief is structural. It crystallises into the load bearing layers of the economy. It moves from private to common to convention to market through identifiable transitions. It is reflexive, so it shapes the reality it describes. And it fails nonlinearly and asymmetrically, which means a system can look strong right up to the moment it is not. The remaining question is why all of this is becoming acute now, visibly, for everyone, at the same time.

6. The Repricing

6.1 What changes when output is abundant

For most of economic history, the binding constraint on most valuable activity was production. Making the thing, performing the analysis, writing the document, composing the design, conducting the audit, drafting the argument: these were scarce because they required scarce human capacity, and scarcity is what gets priced. The belief layer was always there, holding up the money and the institutions, but it was not where most participants felt the constraint, because the bottleneck was upstream of it, in the cost of producing the thing the belief would later be attached to.

That constraint is being removed for a very large class of cognitive output. When a system can produce a competent document, analysis, image, model, or argument at near zero marginal cost and effectively unlimited volume, production stops being the scarce input. This is not a forecast. By 2026 it is the observable condition of several entire categories of work. And the moment production ceases to be scarce, the question of value moves, necessarily and immediately, to the next constraint in the stack. That next constraint is the one this paper has been describing. It is credence: not can this be produced, but can this be believed, and on whose word.

The repricing is straightforward to state and uncomfortable to absorb. When anything can be generated, the generated thing is no longer where the value concentrates. The value concentrates in whatever lets a recipient believe a given output, attribute it to a source, and rely on it. That is the belief layer, surfacing from the foundation of the economy, where it was always doing structural work, to the top of it, where it now becomes the thing that is explicitly scarce, explicitly priced, and explicitly fought over.

It is worth being careful here, because a reader who has seen technological cycles before will reasonably ask whether this is just the usual story of a capability getting cheap and value moving elsewhere, the way mechanisation moved value from muscle to machine operation, or software moved it from manual process to systems design. It is not the usual story, and the difference is exact. In every previous shift, the constraint moved from one form of production to another form of production: from making by hand to making by machine, from computing slowly to computing at scale. The factor that became scarce was still, in each case, a way of producing things. What is different now is that the constraint is not moving to a new form of production at all. It is moving off the production axis entirely, onto the credence axis. The scarce thing is not a better way to make the output. It is the ability to make any output believable and attributable in a world where output itself carries no presumption of authenticity. That is not the next rung on the same ladder. It is a different ladder, and it is the one this paper has been climbing since the first section, because the credence axis is the belief layer, and the belief layer was always the structure. What is new is not that it exists. What is new is that it is now the binding constraint, in the open, priced, for the first time at the scale of the whole economy at once.

6.2 The evidence is already visible

This is not speculative, and the structural facts can be stated without inventing precise figures.

Synthetic media has gone from a curiosity to a majority of new content within a very small number of years. Industry tracking of deepfake incidents records an increase of roughly two orders of magnitude across the 2023 to 2025 window, and credible estimates hold that a large fraction, by some accounts the majority, of online content is now machine generated. The exact percentages are contested and should be treated as reported estimates rather than measured constants. The direction and the order of magnitude are not contested. The default assumption a rational observer now brings to an unattributed artefact has inverted: the prior is no longer that it is genuine unless shown otherwise, but that it could be synthetic unless its provenance can be established.

The response has been an attempt to rebuild a believable signal by technical means. The Coalition for Content Provenance and Authenticity, with backing from major model and device makers, has produced a content credentials standard that attaches a cryptographically signed record of an artefact's origin and edit history. By 2026 it has native support on some flagship devices and verification surfaces on some platforms. It is a serious effort and it is instructive precisely in its limits. Its own documentation and independent analysis are explicit that provenance certifies history, not truth: a valid record establishes where a file came from and what was done to it, not that the scene it depicts is real or that the claim attached to it is honest. The records are routinely stripped when content is recompressed by platforms that do not preserve them. And technically valid manifests can, as has been publicly demonstrated, be attached to forged content.

Read carefully, this is the thesis of the paper restated by the engineering itself. The most rigorous available attempt to mechanise trust does not eliminate the need for belief. It relocates it. The question is no longer "do I believe this image" but "do I believe the signer, do I believe the chain, do I believe the institution that vouches for the chain." The recursion did not disappear when it met cryptography. It moved up one level and waited. There is no terminating layer at which mechanism replaces credence. There is only the question of which institution's word the population is willing to treat as the focal point. That is a belief question, in exactly Schelling's and Lewis's sense, and it is now the central economic question of the era rather than a background condition of it.

6.3 Regulation as a forced focal point

Regulators have arrived at the same place from the opposite direction. The European Union's AI Act, whose transparency obligations apply from August 2026, requires that people be informed when they are interacting with an AI system or are exposed to AI generated or manipulated content, and the harmonised technical standards that will operationalise it are being drawn up through the European standards bodies, with the first expected to begin appearing in 2026. Strip the legal language and the function is clear. The law is an attempt to legislate a believable signal back into a system whose output has stopped authenticating itself. It is a public effort to manufacture a focal point, a place where the population's expectations about what can be relied upon can re converge, because the spontaneous focal points that culture used to supply have been overwhelmed by abundance.

Whether that effort succeeds is, fittingly, a belief problem and not only a drafting problem. A standard becomes load bearing not when it is published but when enough participants believe enough other participants will treat it as the reference, which is the convention to market transition of section 4 operating on regulation itself. The literature on the legitimacy of these standards is openly unsettled, which is the honest state of an infrastructure that is being built in public while everyone watches to see whether everyone else will treat it as binding. The standard that is believed to be the standard will become the standard. The contest underway in 2026 is, at its core, a contest over which belief about reliability the world will crystallise around.

6.4 The scarce input is verified human signal

There is a second, quieter consequence, and it sharpens the first. Systems that learn from data degrade when they are trained recursively on their own kind of output; the synthetic feeding the synthetic loses contact with the thing it was supposed to represent. This phenomenon, sometimes called model collapse, has a precise economic implication. It makes verified human originated signal, data that can be attributed to a real source acting in good faith, not abundant but increasingly scarce, and therefore increasingly valuable, exactly as the abundant synthetic output becomes increasingly cheap. The economy is, in other words, simultaneously driving the price of generated output towards zero and driving the price of believable provenance towards the centre of value. Those are the same movement seen from two sides. They are the repricing this paper named, now visible in the most material place an economy keeps score: in what people are willing to pay for.

To summarise the inflection in one line. For most of history the scarce, priced, and contested layer of the economy was production, and belief did its load bearing work silently at the foundation. When production stops being scarce, the foundation becomes the frontier. The belief layer does not become important in the AI era. It was always the structure. It becomes visible, scarce, and priced, which is a different and more consequential thing.

7. The Substrate Beneath the Argument

7.1 What the argument has been pointing at

A paper like this could end at section 6 and be complete as analysis. We continue, briefly and without changing register, because the analysis has been describing something, and it is more honest to name it than to leave the reader to assemble it.

If belief is the load bearing layer of the economy, and if the AI era removes the old constraint and pushes that layer to the surface where it becomes scarce and priced, then a specific kind of thing acquires foundational importance. Not a product. A substrate. The layer at which belief is held, identity is anchored, and trust is made continuous across systems that can otherwise generate anything and attribute it to no one. Above it sits the orchestration layer, the part that produces, routes, and executes, and that is precisely the part abundance is making cheap. Beneath it has to sit something the orchestration layer cannot supply about itself: a durable answer to whose word this is, what it has consistently done, and why it can be relied upon. Call that the identity layer over the orchestration layer. The orchestration layer is becoming a commodity. The identity layer is becoming the structure.

This is not a prediction about a market that might emerge. It is a description of where the load already sits. Every institution that has weathered a long time has, knowingly or not, been operating an identity layer: an accumulated, externally legible, slow to build and fast to lose stock of believable consistency that let counterparties rely on it without re verifying it each time. What the AI era does is make that layer no longer optional and no longer implicit. When production is free and provenance is the constraint, the institutions that endure will be the ones that treated belief as infrastructure to be engineered and maintained, rather than as a soft factor to be assumed.

7.2 Roots, not features

There is a useful distinction between building on top of a system and building underneath it. Most activity in any era builds on top: products, services, applications, the visible layer that rides on infrastructure taken for granted. A smaller and more consequential kind of work builds underneath: the standards, the trust mechanisms, the units of account, the registries, the things that later activity has no choice but to route through. Section 4 made the point in the abstract. Whoever shapes the convention that everyone else's expectations settle onto has built the riverbed rather than competed in the river.

The reason this matters for the present moment is timing. Infrastructure of belief hardens. Before a standard sets, it is contestable and the cost of shaping it is the cost of an argument. After it sets, it is a focal point held in place by everyone's expectation of everyone else, and the cost of changing it is the cost of re coordinating a whole population. The roots of the AI era's trust and identity infrastructure are forming in exactly this window: provenance standards, certification regimes, the legal definition of a reliable signal, the question of which institution's word becomes the reference. Insurance and certification are not peripheral to this. They are two of the load bearing roots, because they are the mechanisms by which a society converts a private belief about reliability into a transferable, priced, collectively recognised one. They are belief, crystallised on purpose, into infrastructure.

Future Proof is the name, in this paper, for the body of practice that has been operating at that root level: treating belief, identity, and continuity of trust not as outputs of a system but as the substrate a system stands on, and doing so already, in present tense, while the standards are still soft enough to shape. It is not introduced here as a proposition. It is referenced the way one references an institution that is simply assumed to be there, because the argument did not arrive at it. The argument was about it from the first section. A reader who has followed the mechanism this far will recognise that the thing the analysis keeps describing, the layer that holds belief while everything above it is automated, is not a future opportunity. It is where the foundation already is, and the only open question is who was already there before it hardened.

7.3 The ecosystem property

One last structural note completes the frame. Belief infrastructure has an unusual economic property that extractive infrastructure does not. It is regenerative when it is used well and degrading when it is used to extract. Every honoured promise slightly increases the stock of trust available to the next transaction. Every dishonoured one, made common knowledge, decreases it for everyone, not only for the parties involved, because what degrades is the population's generalised expectation that promises of that kind are kept. This means a system built on belief has an incentive structure unavailable to a system built on extraction: the way to make the substrate more valuable is to keep widening the circle of honoured expectation rather than to mine it. An ecosystem that regenerates the belief it runs on compounds. One that extracts from it spends a stock that is far more expensive to rebuild than it ever was to keep. This is not an ethical embellishment on the analysis. It is the analysis. Belief, uniquely among load bearing materials, grows under correct use and fails under extraction, and any institution intending to last in an era where belief is the scarce layer has to be built to the first property and against the second.

8. Implications

The analysis is general. Its consequences are specific to where you stand in relation to the belief layer. We address four positions directly, in the substrate register, without a single recommendation that depends on selling anything.

For institutions

An institution's most valuable asset is almost certainly not on its balance sheet, and the AI era is about to make that fact unignorable. It is the accumulated, externally legible belief that the institution will behave the way it has always behaved. Treat that asset the way a serious operator treats critical infrastructure: inventory it, understand its dependencies, monitor it knowing that its stability and its fragility look identical until the transition, and never read "no visible problems" as "no problem," because section 5 explains precisely why that reading is an error. The institutions that endure the repricing will be the ones that moved belief from the soft column to the engineering column before they were forced to, and that understood that being inside the trust standards while they are still being written is categorically different from complying with them after they have set.

For investors

Most asset valuation already prices belief and most valuation language pretends it does not. A multiple is a crystallised belief about a future that has not happened, sustained reflexively by other investors holding the same belief and aware that they do. The practical consequence of taking this paper seriously is not cynicism about value; it is precision about what is actually being underwritten. In an era where output is abundant and provenance is scarce, the durable returns will concentrate not in what is produced but in what makes produced things believable and attributable: the standards, the verification layers, the identity infrastructure, the institutions positioned at the root of a hardening system rather than competing in the flow above it. The reflexive nature of belief cuts both ways for a portfolio. It is why narratives can sustain valuations longer than fundamentals justify, and it is why the asymmetry of section 5 means the unwinding, when it comes, is faster and more total than the model expected. An investor who has internalised the thin film does not get more nervous. They get better timed.

For operators

If you build, you have a choice the AI era is making consequential: build on top of the belief infrastructure, or build underneath it. Building on top is most of what gets built and it rides on conventions taken as given. Building underneath means shaping the convention itself, the focal point others' expectations settle onto, in the narrow window before it hardens. The window is the point. Before a standard sets, the cost of shaping it is the cost of an argument made well and early. After it sets, the cost of changing it is the cost of re coordinating an entire population's expectations of each other, which is to say it is effectively closed. An operator who understands belief as infrastructure spends disproportionate effort on the layer that everything else will be forced to route through, and spends it while that layer is still soft.

For the people inside these systems

The repricing has a human edge that the structural language can obscure. When output is abundant, what a person is believed to be, consistently and over time, becomes more economically load bearing than what they can produce in any given hour, because production is the part that is no longer scarce. This is not a comfortable adjustment and it should not be described as one. But it has a direction worth naming plainly. A system that treats the people inside it as production capacity to be drawn down will degrade exactly the belief asset that the era is making scarce, in the same nonlinear and asymmetric way section 5 describes. A system that builds and keeps the trust of the people inside it is, by the mechanics of this paper rather than by sentiment, accumulating the one form of capital that is becoming structurally more valuable rather than less. The humane design and the durable design are, in this specific era, the same design. That is not a hope. It is what the analysis says.

9. Coda

Every age has a thing it mistakes for the structure and a thing it mistakes for the weather. For a long time the structure was thought to be the visible production: the output, the throughput, the made thing, the number on the page. Belief was the weather, the soft factor, the thing that passed over the structure and was occasionally inconvenient and was never quite respectable enough to put in the model.

The argument of this paper is that the age had it inverted. The made thing was always riding on the belief, the way a price rides on a currency that rides on a promise that rides on the expectation that strangers will keep honouring it. Money proved it. Institutions proved it. Authority, property, standards, and trust each proved it again at a larger scale. The only reason the inversion was tolerable for so long is that production was scarce enough to feel like the constraint, so the foundation could do its load bearing work in silence and go unnamed.

That silence is ending, not because anyone argued the case but because abundance removed the cover. When the made thing costs almost nothing and can be produced without limit, it stops being where the value is, and the value returns, visibly and all at once, to the layer it never actually left. What can be believed. Whom you can believe. Whose word becomes the place a whole population's expectations agree to meet. Those were always the structural questions. They are about to be the priced ones.

The institutions that last will not be the ones that produced the most in the years when producing was easy. They will be the ones that were already, quietly, treating belief as infrastructure: building underneath the system rather than on top of it, in the short window while the roots were still soft, holding the layer that holds everything else. That window is open now and it does not stay open. Infrastructure of belief is contestable until it hardens, and then it is the riverbed, and everything that comes after has no choice but to flow through it.

Build the layer that holds belief, while belief is still soft enough to shape.

References and Notes

The following are real and verifiable. Where a figure is reported by industry or press rather than independently measured, that is stated in the body and the figure is used only at the level of order of magnitude and direction.

J. R. Searle, The Construction of Social Reality (Free Press, 1995); Making the Social World: The Structure of Human Civilization (Oxford University Press, 2010); "Social ontology: Some basic principles," Anthropological Theory 6(1), 2006.
G. Ingham, The Nature of Money (Polity, 2004).
J. M. Keynes, The General Theory of Employment, Interest and Money (Macmillan, 1936), especially Chapter 12.
G. Soros, The Alchemy of Finance (Simon and Schuster, 1987), and Soros's later writing on reflexivity and the human uncertainty principle.
D. W. Diamond and P. H. Dybvig, "Bank Runs, Deposit Insurance, and Liquidity," Journal of Political Economy 91(3), 1983. Diamond and Dybvig were awarded a share of the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel in 2022, with Ben Bernanke.
D. C. North, Institutions, Institutional Change and Economic Performance (Cambridge University Press, 1990).
M. Weber, Economy and Society (1922; University of California Press edition, 1978), on the three pure types of legitimate domination.
K. J. Arrow, The Limits of Organization (W. W. Norton, 1974), on trust as a lubricant of social systems.
F. Fukuyama, Trust: The Social Virtues and the Creation of Prosperity (Free Press, 1995).
P. J. Zak and S. Knack, "Trust and Growth," The Economic Journal 111(470), 2001.
T. C. Schelling, The Strategy of Conflict (Harvard University Press, 1960), on focal points.
D. K. Lewis, Convention: A Philosophical Study (Harvard University Press, 1969), on common knowledge and the structure of convention.
R. J. Shiller, Narrative Economics: How Stories Go Viral and Drive Major Economic Events (Princeton University Press, 2019), on contagious narrative as an economic force.
Coalition for Content Provenance and Authenticity (C2PA), Content Credentials technical specification, public at spec.c2pa.org; Content Authenticity Initiative, reporting on the state of content authenticity (2026). The limitation that provenance certifies history rather than truth, that manifests are stripped on recompression, and that valid manifests can be attached to forged content is documented in C2PA materials and independent technical analysis.
Regulation (EU) 2024/1689 of the European Parliament and of the Council (the Artificial Intelligence Act); European Commission, Shaping Europe's digital future, AI Act standardisation and implementation timeline, on the application of transparency obligations from 2 August 2026 and on harmonised standards developed through CEN and CENELEC Joint Technical Committee 21, with first standards expected in 2026. The contested legitimacy of these standards is discussed in the standardisation literature.

Note on synthetic media figures: orders of magnitude for the growth of deepfake incidents across 2023 to 2025 and estimates that a majority of online content may be machine generated by 2026 are drawn from content authenticity and digital provenance industry reporting current to 2026. They are used in this paper only directionally and at the level of order of magnitude, consistent with the paper's own rule against cementing precise figures that cannot be independently grounded.

Future Proof Intelligence . Research . No. V . MMXXVI