• The claim
• Why the incumbent can’t serve
• The same shape, four times
• The pattern
• The contemporary instance
• What divergence does and doesn’t mean
• Two honest objections
• What this means
• Where to go from here
• Sources and further reading

Why every new economy has to build its own money

In the 1950s, the Soviet Union had a problem with its dollars.

It held them — reserves it needed for trade — but it held them in the currency of its principal adversary. If tensions rose, the reasoning went, the United States could freeze or seize Soviet dollar deposits sitting in American banks. So the Soviets moved the money somewhere Washington couldn’t reach: London, where banks were willing to take dollar deposits without running them through the US regulatory system.

Those deposits compounded. By the 1960s they had grown into an entire parallel dollar-banking system — the eurodollar market — operating completely outside the Federal Reserve’s perimeter: no reserve requirements, no deposit insurance, no interest-rate ceilings, no US supervision.

Here is the part that matters. American banks could not have hosted those deposits. The very things that made a US bank trustworthy — Fed oversight, FDIC backing, regulatory discipline — were precisely the things the eurodollar market existed to escape. A US bank that pulled the deposits inside the perimeter would have lost them to a London bank outside it. To compete, it would have had to become an offshore bank — at which point it would have stopped being a US bank.

So it didn’t. The activity routed around the incumbent and built its own infrastructure. By the 1970s the eurodollar market was the dominant dollar-funding rail on earth, and it still is.

That is not, in the end, a story about banking. It’s a story about a pattern — one that has repeated across centuries and across domains, and that is repeating right now around a question almost nobody framed this way until recently: what kind of money does an autonomous AI agent actually use?

This essay is about that pattern, and the structural claim underneath it. Call it the Independence Doctrine. The full version, kept current, lives at bitcoineconomy.ai; this is the argument’s spine.

§The claim

The claim is sharp: emergent parallel economies must structurally diverge from incumbents to succeed. The dominant system cannot give the new economy what it needs without ceasing to be the dominant system. The result is not a friendlier, more accommodating incumbent. It’s a parallel system — different properties, different participants, different governance — running alongside the old one and touching it only at narrow bridge points.

And the claim is structural, not ideological. I’m not arguing that incumbents are villains, that the new economy is virtuous, or that it’s destined to win. It’s an architectural observation: when a new activity requires properties the dominant infrastructure structurally cannot provide, the activity routes around it and builds its own substrate. The AI economy on Bitcoin is the present-day instance of a very old pattern.

§Why the incumbent can’t serve

The mechanism is precise enough to state plainly.

An institution becomes dominant through a bundle of properties — scale, regulatory accommodation, predictability, trust enforcement, integration with everything around it. That bundle is what makes it dominant. It’s also exactly what its existing customers chose it for.

When a new activity shows up needing different properties — settlement without permission, routing without oversight, payments below the fee floor — the incumbent faces a choice that isn’t really a choice. To serve the new activity, it would have to relax the bundle that defines it. Relaxing the bundle loses the customers who chose it for that bundle. It cannot serve both at once; the properties one side needs are mutually exclusive with the properties the other side selected.

So it doesn’t serve the new activity. It can’t. And this is not a story about resistance or bad faith. Bank executives are not refusing to serve autonomous agents out of malice; stablecoin issuers are not declining censorship-resistance out of ideology; central-bank-currency architects are not preserving freeze capability out of authoritarianism. Each is behaving consistently with the properties that define it — properties its regulators, customers, and shareholders require. The architecture of the legacy stack is the price of its dominance. Dismantle the architecture and you dismantle the dominance.

So the new activity builds its own rails, and the properties the incumbent couldn’t provide become the foundation the new system is built on.

§The same shape, four times

You’ve already seen one instance: the eurodollars. Here are three more, chosen for range — technological, political, logistical — so the shape reads as structural rather than as a quirk of finance.

The open internet vs. the walled gardens. Through the early 1990s, going online meant AOL, CompuServe, Prodigy — curated services behind a subscription, a proprietary client, and a central moderation layer. AOL alone peaked above 20 million paying subscribers. The properties that sold the gardens — curation, support, a predictable interface — were the business. The open internet had the inverse properties: anyone could publish, anyone could host, no authority could veto a domain. To early users those looked like bugs. But they were exactly what new applications required, and the gardens couldn’t adopt them without ceasing to be gardens. AOL built a browser; it could never become the browser without surrendering the curation and account control that defined its subscriber relationship. By 2005 the open web had won, and the surviving gardens were reduced to content layered on top of it.

Samizdat vs. the Soviet state press. The state press was dominant by every measure — universal distribution, official authorization, professional production. Those were also the properties that made it an instrument of state ideology, which structurally precluded publishing dissident work. A state press that printed The Gulag Archipelago would have stopped being a state press. So the activity routed around it: samizdat, the hand-to-hand circulation of typewritten and carbon-copied texts. Amateur, illegal, tiny print runs — and the only viable rail for the activity it served. It never rivaled the state press in volume; for its activity, the state press was simply irrelevant.

Private couriers vs. the postal monopoly. The US Postal Service held a legal monopoly on first-class mail, built on universal coverage and low, batch-processed, time-indeterminate delivery. Those properties made it dominant — and made overnight, tracked, signature-guaranteed delivery structurally impossible for it. By the late 1960s, commerce needed exactly that. FedEx launched in 1971; UPS and DHL followed with the same hub-and-spoke architecture. USPS built Express Mail, but it could never become the dominant overnight rail without becoming a private courier — and abandoning the universal-coverage charter its monopoly required. Today the two share customers but not architecture: USPS still moves most first-class mail, private couriers handle nearly all overnight commerce.

§The pattern

Across all four, the same shape recurs:

• The dominant system has a property bundle that is its institutional identity.
• A new activity emerges needing properties outside that bundle.
• The incumbent can’t adapt without ceasing to be itself — every one of these incumbents had the incentive and the capability to capture the new activity, and not one of them did.
• A parallel system forms with the inverse properties.
• The two coexist through narrow bridges that never merge the architectures.
• The parallel system becomes dominant for its activity — not necessarily overall (USPS still moves most first-class mail), but for the thing it formed around.

It doesn’t depend on motives or ideology or institutional detail. It depends only on a structural property-mismatch between what the incumbent provides and what the new activity requires. Where the mismatch is structural, divergence is the predicted outcome.

§The contemporary instance

Now the present case. Each element maps cleanly.

The dominant system is the incumbent payment stack — banks, card networks, regulated stablecoin issuers, central-bank rails — defined by regulatory accommodation, identity intermediation, freeze capability, and central coordination. That bundle is what makes it work for human payments, and what regulators require it to keep.

The new activity is autonomous AI agents transacting at machine tempo, which need four properties at once: KYC-free access, censorship-resistance against an intermediary freeze, sub-cent settlement, and machine-tempo speed. Each is structurally absent from at least part of the incumbent stack.

The incumbent can’t adapt. Banks can’t go KYC-free without losing their charters. Stablecoin issuers can’t refuse freeze orders without losing regulated status. Central-bank-currency architects can’t disable monitoring without contradicting the reason the currency was issued. Card networks can’t drop per-transaction fees without abandoning the interchange model that funds them. Each accommodation would unwind the identity that makes the participant dominant.

The parallel infrastructure already exists — Bitcoin’s base layer for settlement, the Lightning Network for sub-cent payments at machine tempo, ecash systems (Cashu, Fedimint) for the small and the private, and the agent-integration primitives on top (L402, Nostr Wallet Connect, MCP servers, the Lightning Agent Toolkit). The inverse property bundle, deployed today.

And the substrate selection has already started, before scale. In March 2026 the Bitcoin Policy Institute ran 36 frontier models through 9,072 neutral scenarios with no leading prompts. Bitcoin was the top overall monetary choice in 48.3% of responses and the preferred store of value in 79.1%, with more than nine in ten responses favoring digitally-native money over fiat. The choice is being made quietly, in inference, long before agent-driven money flow is large enough to register macroeconomically.

There’s even a field test. Faced with the agent economy, the incumbents did not adopt Bitcoin’s properties — they shipped parallel agent-payment rails that preserve their own (identity-bound wallets, freeze-capable issuers, regulated processors), aimed at the use cases that don’t require what they withhold. That they built a parallel stack rather than adapting toward Bitcoin’s is the doctrine confirmed, not refuted. The live contest this provoked — the competing rails, the “automation, not agency” wedge, the integration objection in full — is the subject of Border Skirmishes.

§What divergence does and doesn’t mean

The doctrine is precise enough to mislead if read carelessly. Three clarifications.

Divergence is not separatism. The parallel system doesn’t have to reject the incumbent to succeed. Bridges are expected and welcome — they make the new system more useful to people who also live in the old one. What divergence requires is architectural independence: the new system’s properties cannot depend on the old one’s permission. Bitcoin can interact with banks through custodians; it cannot depend on a bank’s approval to settle a transaction.

Divergence is not zero-sum. The eurodollar market didn’t destroy US domestic banking; it added a substrate for a new activity. The open internet didn’t destroy commerce; it hosted vastly more of it than the gardens ever could. The agent economy on Bitcoin won’t destroy human commerce on bank rails; it adds a substrate for an activity those rails were never architecturally suited to host.

Divergence is not a prediction about timing. The doctrine predicts the shape of the outcome, not its speed or size. Eurodollars took two decades to dominate global dollar funding; the open internet took about one to eclipse the gardens; samizdat stayed niche throughout the Soviet period; couriers became dominant for overnight delivery within a decade of FedEx. The doctrine predicts shape; the record resolves magnitude over time.

§Two honest objections

Two objections deserve a direct answer. A third — that regulated stablecoin rails simply absorb the bulk of agent payments and Bitcoin stays a niche — is the live contest, engaged at length in Border Skirmishes.

The four analogues might not generalize. Four hand-picked cases invite over-fitting, and there are real cases where incumbents did adapt — record labels to streaming, newspapers to digital subscriptions, card networks to mobile wallets. The response is that the doctrine’s claim is conditional. It does not predict that incumbents always fail to adapt; it predicts they fail when the new activity requires properties their institutional identity structurally precludes. Streaming, digital news, and mobile wallets all let the incumbent’s identity survive — labels still license catalogs, newspapers still produce journalism, card networks still intermediate identity. The four analogues did not allow that survival. The AI economy on Bitcoin sits in the latter category, and the burden is on critics to show otherwise.

The doctrine might be self-defeating. State it clearly and you hand regulators and incumbents a roadmap to preempt the divergence before the parallel system is resilient. That risk is real, and worth naming with eyes open. But the argument does not depend on stealth; it depends on the property mismatch. Regulators can pressure Bitcoin payment infrastructure, but they cannot make banks censorship-resistant or make issuers refuse freeze orders without restructuring the institutions themselves. The prediction holds even under hostile conditions — the bridges simply get narrower and the divergence sharper.

§What this means

If you’re building agent infrastructure, the takeaway is concrete: keep the protocol layer — Bitcoin, Lightning, Cashu, Fedimint — permissionless and censorship-resistant, and don’t bake compliance into it. Build compliance at the gateway instead, at the custodian or on-ramp that mediates between the new stack and the old one. Two layers, two different jobs. Plan for narrow bridges, not unification: some agent commerce really must touch the legacy stack, but the protocol layer shouldn’t be compromised to make those crossings cheaper. The separation is the point.

And if you’re just reading to see where this goes, recognize what you’re looking at. The substrate for the agent economy is not “to be decided.” It’s being decided right now — in the inference behavior of frontier models, in the toolkits shipping today, in the architectural choices of every team building agent-payment infrastructure. And it’s being decided in favor of the road that has no walls in the places where agents need none.

The doctrine doesn’t say the legacy stack is doomed. It says the legacy stack is structurally unsuited to host the AI economy, and that the AI economy will therefore form alongside it — the way the eurodollar did, the way the open internet did, the way samizdat and the couriers did. Each time, the incumbent couldn’t adapt. Each time, a parallel system formed. This is the present-day instance of the pattern.

§Where to go from here

This essay is one room of a larger argument that lives — kept current, at whatever depth you want it — at bitcoineconomy.ai.

• The Case — the substrate question itself: the four constraints, why the legacy stack fails them, why Bitcoin on Lightning is the only deployed system that meets all four, and why a brand-new “agent-coin” can’t substitute.
• The Stack — the architecture the parallel system runs on, at engineer depth.
• Border Skirmishes — the live test of this doctrine: the incumbents’ competing rails and the integration objection in full.
• Field Notes — what’s true right now, week to week, as the deployment record accumulates.

The canonical version of this essay lives at bitcoineconomy.ai/independence-doctrine.

§Sources and further reading

• Catherine R. Schenk, The Origins of the Eurodollar Market in London: 1955–1963 (1998) — the canonical account of eurodollar emergence outside Fed supervision.
• Tim Wu, The Master Switch: The Rise and Fall of Information Empires (2010) — the open-internet vs. walled-garden dynamic.
• H. Gordon Skilling, Samizdat and an Independent Society in Central and Eastern Europe (1989) — the scholarly treatment of samizdat networks.
• Roger Frock, Changing How the World Does Business: FedEx’s Incredible Journey to Success (2006) — the founding of overnight delivery.
• Bitcoin Policy Institute, Study: AI Models Overwhelmingly Prefer Bitcoin and Digital-Native Money Over Traditional Fiat (March 2026) — the empirical anchor for substrate preference. btcpolicy.org · canonical study site moneyforai.org.