• The Institutions Were Built for a World That Ended
  • The gap, in numbers
  • Why institutions are slow, and why that was once a feature
  • The specific mechanics of the mismatch
  • What trust in money actually rests on
  • Bitcoin’s actual claim, stated without romance
  • The three functions, examined honestly
  • The strongest argument against all of this
  • A future that could plausibly happen
  • What I would actually watch
  • The uncomfortable part

§The Institutions Were Built for a World That Ended

Something is being decided right now, in offices and server halls, that our institutions have no procedure for deciding. Not because they are captured, or corrupt, or staffed by fools. It is because they were designed to be slow, and the thing they must now govern gets faster every quarter.

§The gap, in numbers

In January 2026, the Edelman Trust Barometer reported that seven in ten people are unwilling or hesitant to trust institutions they once took for granted. The trust gap between high and low income respondents has widened from six points in 2012 to fifteen. Among the forces destroying trust, inflation ranks first at 54 percent, misinformation second at 50 percent, and the growing use of generative AI fifth at 34 percent.

Trust in national government leaders fell sixteen points in a single year. Trust in major news organisations fell eleven. For the second consecutive year, the developed economies sit at the very bottom of the index: Japan at 38, France at 42, Germany and the United Kingdom at 44.

These are not numbers about AI. They are numbers about the vessels into which AI is being poured.

In the same season, Stanford’s AI Index reported that industry produced over ninety percent of notable frontier models in 2025, that several of those models now meet or exceed human baselines on PhD-level science questions, and that generative AI reached fifty-three percent population adoption within three years. That is faster than the personal computer, faster than the internet at the same stage. United States private AI investment reached 285.9 billion dollars in 2025, more than twenty-three times the figure for China and fourteen times Europe’s.

And then this, from the same report: documented AI incidents rose to 362, up from 233 the year before, while reporting on responsible-AI benchmarks remains spotty and safety benchmarks lag capability.

Read those two datasets together and the shape becomes clear. The technology is compounding. The trust is eroding. And the oversight is falling behind both.

§Why institutions are slow, and why that was once a feature

A parliament takes years to pass a law. A regulator takes months to issue a rule. A court takes years to decide a case. A central bank meets eight times a year.

We call this slowness, and we mean it as an insult. It is not. Slowness is the mechanism by which a society prevents any single actor (including a popular one, including a frightened one) from rewriting the rules overnight. The delay is the protection. It is why a phone call from a president cannot change the law on Tuesday, and why a panic cannot liquidate the constitution by Friday.

Every institution we have inherited carries this design. Double-entry bookkeeping, limited liability, central banking, the limited state, the independent judiciary, the four-year electoral cycle: all of them are brakes. They were built to be slower than the moods of men.

That worked while the world they governed was also slow. A factory took years to build. A new product took a decade to reach a market. A currency crisis unfolded over weeks. A bank run required people physically queuing in the street.

None of that is true anymore. And here is the part that is genuinely new: the thing now moving faster than the institutions is not a market or a competitor. It is cognition itself.

§The specific mechanics of the mismatch

It is easy to say “AI moves fast, regulation moves slow” and feel that one has said something. It is harder, and more useful, to name the exact places where the mismatch bites.

First: the unit of regulation is the model, but the unit of risk is the deployment. A regulator can license a model. It cannot license the countless uses to which an agent puts that model by Thursday. The EU AI Act, the most comprehensive attempt anywhere, governs providers and deployers. It does not and cannot govern the combinatorial space in which a general-purpose reasoning system meets a specific workflow inside a company nobody has audited.

Second: liability assumes a human in the loop. Tort law asks who was negligent. When an agent autonomously sends a payment, drafts a filing, negotiates a contract, or writes the code that runs a factory, the chain of causation becomes diffuse rather than merely long. The Stanford report notes agents jumped from twelve to sixty-six percent task success on real computer tasks in a single year. At that trajectory, the “human in the loop” becomes a signature on a document nobody read.

Third: the identity layer is broken and nobody has replaced it. Every institution in the developed world, from banks to welfare states, tax authorities and elections, rests on the assumption that a person can be identified and that a document can be verified. Both assumptions are collapsing simultaneously. A voice can be cloned from three seconds of audio. A face can be synthesised. A signature is a rendering. The institutions built cheque-clearing and passport control on the assumption that forgery required skill; it now requires a prompt.

Fourth: the labour market adjustment has moved from forecast to measurement. Goldman Sachs puts a nine percent displacement at fifteen million workers. The ECB reports that the share of European workers using AI on the job doubled from twenty-six to fifty-two percent in two years. The S&P Global survey found negative employment impact already reported in the past twelve months. Not projected. Reported.. Tax systems, pension systems, and social insurance were designed around a model of employment that is being hollowed out from the middle, and none of them has a mechanism to adjust at this speed.

Fifth, and least discussed: the state’s monopoly on money is being technically and politically questioned at the same time. That deserves its own section.

§What trust in money actually rests on

Strip away the ritual and a currency is a promise, held in place by an institution that can be trusted to keep it. The dollar is a promise by the Federal Reserve and the Treasury. The euro is a promise by the ECB and the member states behind it. Both are promises about quantity: that the issuer will not debase beyond some tolerable rate — and both depend entirely on the credibility of the issuer.

For most of the postwar period, that credibility was not seriously questioned, because the alternative was worse and because the institutions delivered. That condition has been eroding for two decades: 2008, then the long zero-rate era, then the 2022 inflation shock, then the fastest rate-hiking cycle in forty years, then a period in which several developed-world governments have run large structural deficits with no political constituency for correction.

The Edelman data shows the result. Trust in government leaders is down sixteen points. In the countries at the bottom of the index (Japan, France, Germany, the UK), the institution that issues the money is not held in markedly higher regard than the institution that collects the tax.

Into that gap, something structurally different has arrived.

§Bitcoin’s actual claim, stated without romance

Bitcoin is not going to replace the dollar, and anyone who tells you otherwise is selling something. It is too volatile to be a unit of account, too slow at the base layer to clear global retail payments, and too politically exposed to serve as the settlement rail of a state that might need capital controls.

But Bitcoin makes one claim that no other asset makes, and that claim becomes more valuable precisely as institutions become less trusted.

Its supply schedule is a rule, not a policy.

There is no committee. At block 974,519 there are 20,105,365 bitcoin in existence; the network pays 3.125 per block, halving again in 2028. No central banker can decide, at a weekend emergency meeting, that this number should be different. No finance ministry can instruct it to buy its own bonds. It has no issuer who can be pressured, no board that can be replaced, no jurisdiction that can freeze it at the protocol level.

For most of Bitcoin’s history this was a curiosity: a technically elegant property with no institutional use case, because institutions were functioning and didn’t need a neutral settlement layer.

That has changed, and the change is visible in policy, not just in price. The United States became the first developed economy to formally adopt Bitcoin as a reserve asset. National bitcoin reserves, which were a fringe proposal in 2021, are now a live instrument of statecraft in a growing list of countries. And the reason is not ideological enthusiasm. It is this: a reserve asset that no other state can sanction, freeze, or inflate is useful to a state that no longer trusts the other states.

That is a very different argument from “number go up.” It is an argument about the architecture of trust in a fragmenting world.

§The three functions, examined honestly

If Bitcoin plays a role in the coming decade, it will be in one or more of three capacities. Each has a different probability and a different set of obstacles.

As a reserve asset. This is already happening, but slowly and unevenly. The obstacle is not technology but accounting: a reserve asset that can lose a third of its value within a quarter is difficult to hold under existing prudential rules. Expect sovereign holdings to remain small, and to be held for geopolitical rather than monetary reasons.

As a neutral settlement layer between blocs. This is the most interesting and least discussed. If the world splits into currency blocs with sanction regimes, the rails between them become a political problem. A settlement asset with no sovereign owner is one answer. It is not a comfortable answer for any government, which is exactly why it might be adopted quietly rather than announced.

As a savings instrument for individuals in failing currencies. This is already the dominant real-world use. In Turkey, Argentina, Nigeria, Lebanon, the demand is not ideological. It is a response to watching savings evaporate. That demand is the most durable of the three, because it is rooted in a problem institutions created and have not solved.

And the honest counterweight: Bitcoin’s market capitalisation today is about 1.58 trillion dollars, roughly 5.4 percent of the value of all gold ever mined. It is a young, volatile asset whose price is still driven more by liquidity cycles than by monetary theory. In March 2020 it fell further and faster than equities. In 2022 it fell more than three quarters from its high. Anyone who tells you it is a stable anchor is confusing the rule with the price. The rule is stable. The price is not.

What Bitcoin offers is not stability of price. It offers stability of the rule that governs supply, which is a different and arguably more important thing when the alternative is a rule that can be changed by a phone call.

§The strongest argument against all of this

I want to state the counter-case properly, because the version of this essay that skips it would be propaganda.

The counter-case is that institutions adapt, and always have. The telegraph did not end the state; it produced the modern regulatory agency. The railroad did not end the state; it produced antitrust law. The internet did not end the state; it produced a vast apparatus of data protection, and the state emerged with more surveillance capacity than it had in 1990, not less. On this reading, AI will produce new institutions: AI auditors, model registries, compute accounting, liability regimes for autonomous agents — and the lag will be measured in years, not decades. The institutions will catch up, as they did with every prior general-purpose technology, and the people who wrote books about collapse will look silly.

There is real force in this. The OECD’s Digital Government Outlook finds that nearly all member countries now have an AI-in-government strategy and thirty of thirty-six have an institution responsible for governing it. That is not nothing. Governments are not standing still.

But there is a specific difference this time, and it is not the speed of the technology. It is the speed of the trust decay. The telegraph did not cause a sixteen-point collapse in trust in government in a single year. The railroad did not produce a world in which seven in ten people decline to trust the institutions that govern them. Previous technologies were adopted inside a functioning institutional consensus. This one is arriving as that consensus is failing, which means the institution has to do two things at once: build its capacity and rebuild its legitimacy. Simultaneously is much harder than sequentially.

And there is a second difference: previous technologies were tools. This one is a substitute for judgement, which means every institution that relies on human judgement (courts, regulators, parliaments, central banks) is being asked to delegate part of its core function to a system it cannot audit.

§A future that could plausibly happen

Not a prediction. A scenario, with its assumptions stated.

By 2030, three things have happened. AI agents handle a meaningful share of routine commercial and administrative work, and the labour market has bifurcated: people who direct agents and people who are displaced by them, with a shrinking middle. Governments have passed AI liability laws, but the enforcement is patchy because the technical capacity to audit a model sits in a handful of firms. And the trust index has continued to fall in the developed world, because they are now slower than the world they describe, while their formal authority remains unchanged.

In that world, the demand for something that cannot be inflated or frozen does not come from libertarians. It comes from pension funds in countries whose currency has been debased, from finance ministries diversifying away from assets that can be sanctioned, and from ordinary people who have watched two decades of monetary experiments and drawn conclusions.

Bitcoin’s role in that scenario is not as money. Nobody buys coffee with it. Its role is as the reference point against which the credibility of every other promise is measured, the thing that cannot be printed, sitting there on the other side of the ledger, making every issuer’s restraint visible by comparison. Central banks would hate this. They would also find it useful, in the same way that gold was useful: because it disciplines the alternative, even though nobody transacts with it.

And the scenario has a failure mode, which must be stated. If AI makes verification cheap and pervasive, if every transaction, document, and claim becomes verifiable by default — then the problem Bitcoin solves (trust without an intermediary) becomes less scarce. A world with perfect cryptographic attestation everywhere needs less of a trustless asset. That is the most serious long-term threat to the thesis, and it comes from the same technology that drives the demand.

§What I would actually watch

Not the price. Four indicators:

One: whether AI liability law becomes enforceable. Not passed, but enforceable. If by 2028 there is a working mechanism to assign liability to an autonomous agent’s operator, the institutional adaptation thesis is winning.

Two: whether sovereign reserves grow past token allocations. A token allocation is a signal. A standing allocation is a policy.

Three: whether the trust index stabilises. Edelman’s numbers have fallen for five consecutive years. A flat year would be the first evidence that the decay is cyclical rather than structural.

Four: whether AI makes verification cheap. This is the one that cuts against Bitcoin. Watch the cost of proving that a document, a person, or a claim is authentic. If it collapses to near zero, the case for a trustless settlement layer weakens everywhere.

§The uncomfortable part

The honest conclusion is not that Bitcoin will save us, or that institutions will collapse. It is duller and more specific.

We are asking nineteenth-century institutions to govern twenty-first-century cognition, and we are doing it while public trust in those institutions is at a record low. That is the actual situation. The gap between the speed of the technology and the speed of the institution is not new. But the gap between the speed of the technology and the willingness of people to accept institutional answers is.

In that gap, two things become valuable. One is legibility: systems whose rules can be read and checked by anyone, without asking permission. The other is neutrality: assets and protocols that do not require you to trust the entity that made them.

Bitcoin is an imperfect example of both. It is volatile, energy-hungry, and politically inconvenient. But it is one of the very few pieces of infrastructure built in the last twenty years whose rules are published, verifiable, and unchangeable by their author.

The institutions we have cannot be rebuilt overnight. But the world they were built for has already ended. And in the space between those two facts, something has to hold.

Roger, 3 November 2026. Market data from own measurement, 3 November 2026: bitcoin 78,520.62 USD, supply 20,105,365 BTC, block height 974,519, gold 4,051.20 USD per ounce. Trust and AI figures: Edelman Trust Barometer 2026 (published 18 January 2026); Stanford HAI AI Index Report 2026; OECD Digital Government Outlook 2026; ECB blog, 26 August 2026; Goldman Sachs, 2 July 2026; S&P Global, 2 June 2026; World Gold Council above-ground stock, Q2 2026.