Monetary Foundations & Financial Stability

The agentic economy does not need a new theory of money. It needs a new theory of the plumbing. Software agents will transact at machine speed. They will move value across many uses in milliseconds, in amounts from fractions of a cent to billions of dollars. Through all of this, the central bank's most important tool, the price of money, does not weaken. It sharpens. The money these agents hold is fully reserved in short-term government debt and central-bank balances. So the policy rate reaches the reserve base of the money supply at once and in full, without the lag and slippage of the bank-lending channel. The lever the central bank cares about most stays intact, and arguably grows more potent.

What changes is everything downstream of that lever. The fractional-reserve multiplier, the way a banking system creates velocity by lending the same deposited dollar many times, fades as a transmission channel, because full-reserve money does not multiply. As it fades, the supervisory job shifts. The monetary authority spends less energy inferring the multiplier and more supervising a transparent, machine-run credit market that grows on top of the money. This does not hollow out the mandate. It moves the mandate from a channel that was always opaque and lagged to one that is, for the first time, readable in real time.

The policy agenda is to manage that shift well. It divides into six tasks: engineer the quality of the money's backing and decide who may hold central-bank reserves; build the singleness of money by design, so many issuers still produce one dollar; preserve monetary transmission and manage the transition of bank credit; supervise a credit market that clears in milliseconds; build the prudential brakes, finality rules, and insurance a machine-speed system needs; and decide who receives the income a fixed, non-yielding base unit throws off. Each is tractable. None is trivial. Each is taken up below.

Engineering the reserve-quality ladder and central-bank access

At machine speed, money must hold par on a "no questions asked" basis, or it cannot work. An agent settling thousands of transactions per second cannot stop on each one to check whether the unit it receives will redeem one-for-one. If it has to doubt, it has to price the doubt. Pricing redemption risk on every micro-transaction at machine speed is impossible. So velocity forces the question of what backs the money to the front of the design. In the legacy system that question sat in the background, cushioned by deposit insurance and a lender of last resort that most people never thought about.

The answer the agentic economy converges on inverts the central reflex of fractional-reserve banking. In that system, scale concentrated risk. The larger and more interconnected an institution grew, the more dangerous its leverage, and the more the public balance sheet stood behind it. Here the logic runs the other way. The larger and more systemic a digital dollar becomes, the closer its backing should move toward central-bank money itself. A small issuer might back its money with short-term government debt. A dominant, economy-scale digital dollar should be backed, at the margin, by direct central-bank cash and overnight repos with the central bank. Safety scales with size instead of degrading with it. The bigger the money grows, the safer its backing becomes: the opposite of the dynamic that produced the last century's banking crises.

This delivers central-bank-grade safety without a central-bank digital currency. The money can be private, programmable, internet-native, and competitively issued, while being bankruptcy-remote, fiduciary-chartered, and backed by the safest asset that exists. The state does not need to issue the currency to make it safe. It needs to build the ladder of eligible reserves and decide who may climb it.

Three jobs follow. First, settle reserve-account and master-account access for permitted issuers, so a regulated full-reserve issuer can hold central-bank balances directly rather than route its safety through a commercial bank. Here a real constraint deserves a clear-eyed answer: very large balances of perfectly safe money sitting at the central bank could change funding markets and policy transmission, and that concern is met on its own terms, through holding caps, tiered access, remuneration design, and transition management. Second, define the reserve-eligibility ladder in statute and rule (which assets qualify at which tier, and how eligibility tightens as an issuer grows systemic) instead of discovering it case by case. Third, harmonize across borders. The Bank of England's evolving model weighs how much of an issuer's backing should sit in short-term government debt versus accounts at the central bank, and European payment-systems law already contemplates e-money issuers' access to central-bank balance sheets. Convergence matters because the money is borderless, and regulatory divergence becomes arbitrage. Build the ladder deliberately, and the result is the safest money the system has ever held, issued competitively and at internet scale.

Singleness of money and par-preservation at internet scale

A money economy can tolerate many issuers only if every issuer's dollar stays interchangeable with every other's. The danger of open issuance is precisely that it produces many "dollars" (each a distinct obligation with its own risk) whose prices can drift apart, so a dollar from one source no longer trades at par with a dollar from another. When that happens, fungibility breaks, and money loses the property that lets it serve as a unit of account. Economists call this property the singleness of money. Preserving it is non-negotiable.

The good news is that singleness is not a natural fact to hope for. Policy can manufacture it by design. Three choices, taken together, produce it. A uniform full-reserve requirement removes the credit-risk difference between issuers, so no issuer's dollar is riskier than another's. An enforced guarantee of par redeemability ensures every unit redeems one-for-one on demand, anchoring its price to par by arbitrage. And cross-issuer interoperability (par-settlement rails where any compliant issuer's dollar clears against any other's at one-to-one) makes the dollars interchangeable in practice, not just in theory.

Singleness can be built. The work is to codify uniform full-reserve backing and guaranteed par redemption in law, to build the cross-issuer par-settlement infrastructure that makes interchangeability real, and to pursue the international standards convergence that extends par across borders and currencies. Full-reserve money is, in fact, the only form that holds par for everyone, everywhere, without national backstops that do not reach a borderless system. That is its decisive advantage as the base unit of a global economy, not an incidental one.

Monetary transmission and bank disintermediation

If deposits migrate from fractional-reserve banks into full-reserve digital money, the deposit base the multiplier runs on shrinks, and the multiplier weakens as a policy channel. This is the disintermediation concern at its sharpest, and it deserves a direct answer rather than reassurance.

The answer has two parts, and both point to a central bank whose hand strengthens. First, the price lever stays fully intact and transmits more directly. Because full-reserve money is backed by short-term and overnight instruments, a change in the policy rate flows into the reserve base at once and in full, without waiting for banks to reprice loans and deposits. The transmission is cleaner than the bank-lending channel ever was. Second, and more striking, onchain credit is observable in real time. A central bank has always inferred credit conditions from lagged, aggregated, after-the-fact reports. Now it can watch the credit market directly as it forms. Its role partly shifts from running the multiplier (a channel it never fully saw) to supervising a transparent machine-credit market it can see continuously. Supervision expands; it does not hollow out.

The work follows. Build the real-time observability that turns the onchain market into a live instrument panel for the monetary authority. Structure rate and balance-sheet operations to work independently of the bank-lending channel; open-market operations and balance-sheet interventions, shown at both ordinary and extraordinary scale over the past two decades, never depended on fractional-reserve leverage in the first place. And manage the transition for community lenders deliberately, since they serve relationship-dependent small businesses and under-banked borrowers that machine underwriting reaches last. Holding caps on digital money are the instrument for this: not a permanent feature, but a transitional brake that slows deposit migration while the new credit market reorganizes and extends its reach. Sequenced this way, the transition ends with a central bank that transmits its rate more cleanly and sees its credit markets more clearly than ever before.

Supervising transparent, machine-run credit markets

The credit market that grows on full-reserve money clears in milliseconds. Agents underwrite it, not committees. New roles populate it: agentic underwriters, insurers, oracle providers, credit-pool operators, orchestrators, each agentic in substance but traceable, through the accountability chain, to a real and answerable human. Failure here can come not from a loan officer's misjudgment but from a model's blind spot or a corrupted data feed. This is a different supervisory object than regulators have faced, and it calls for a different posture.

That posture flips supervision from forensic-and-lagged to ex-ante-and-live. Today, supervision means reconstructing what happened from quarterly filings and stress tests run on stale snapshots, and the correlation that mattered in the last crisis stayed invisible until it detonated. In a transparent onchain market, the aggregate exposure graph is observable as it forms: every receivable, every stake, every counterparty link. Selective-disclosure and confidential-computing techniques mean this does not force private books onto a public ledger. A supervisor holds cryptographically enforced, permissioned read access to the system-level picture in real time, while a firm keeps its positions confidential from competitors. Intervention stops being a blunt after-the-fact instrument and becomes a graduated dial: correlation-aware margining and system-wide parameters that tighten automatically as concentration builds.

And the credit reaches the long tail of borrowers who were never rationed by their quality, only by the cost of evaluating them. Credit has always been rationed by the price of underwriting, not the soundness of the borrower. When that cost collapses, an enormous population of sound but previously unbankable borrowers becomes serviceable for the first time. Credit becomes cheaper, more abundant, and safer at once, because the new efficiency comes from better information, not more leverage. This is the credit-side expression of velocity replacing leverage: volume grows on superior underwriting and high turnover, not on the manufacture of risky synthetic dollars.

The work is concrete. Define the regulatory perimeter, separating the systemically relevant actors (those whose books or data feeds can move the whole graph) from the long tail. Adopt a two-tier model: explicit registration and licensing for the systemic actors, standards and supervised self-regulation for the rest. Position central banks alongside capital-markets regulators as joint rule-setters, since the credit market sits at the seam of both mandates. Guard the two-tier model against capture by large agent-operators and against neglect of correlated risk in the long tail, and watch the new structural failure modes: model monoculture, where one dominant underwriting model misprices everywhere at once, and oracle concentration, where a shared data feed becomes a single point of failure. And build the supervisory technical capability: the quantitative and engineering skill to read a live exposure graph, because observability is not judgment, and a live map helps only a supervisor who can read it. Standardize the real-time read-access protocols that make permissioned supervision interoperable across authorities, since the market is borderless while supervisors are national. For the first time, those responsible for the system's stability would work from a live, verifiable picture of it, and could intervene gradually rather than bluntly. That is a materially more resilient foundation than the one in place today.

Deterministic finality, the prudential brake, insurance, and the backstop question

Agents need deterministic, sub-second finality: settled must mean settled, now, with no chance that a confirmed transaction is later un-recognized. Older decentralized designs offered only probabilistic finality (wait some confirmations and a reversal becomes unlikely) which agents transacting at extreme velocity cannot operate on. A chain engineered for deterministic finality can meet the requirement; older designs cannot. Yet people also want reversibility: refunds, fraud protection, the ability to undo a mistake. And a credit market clearing in milliseconds needs brakes that act faster than any committee can convene. The answer to each is architectural, not rhetorical.

On finality and reversibility, the move is the one the internet already made: layer reliable protocols over a simple base. Keep the base money deterministically final, and build reversibility as optional protocols on top (time-adjudicated escrow, refund pools, insurance against those pools) rather than welding it into the money itself. Reversibility baked into the base unit would destroy the no-questions-asked property and force every agent to price unwind-risk on every transaction. Pushed to the edges as composable layers, the same protections become available without compromising the core. A final settlement is a guarantee an agent can build on precisely because it never has to think about it again.

On brakes, the key instrument is not an off-switch but a dial: correlation-aware margining that raises the cost of crowding into the same demand source, model, oracle, or compute provider as that crowding forms, so risk grows steadily more expensive instead of hitting a sudden wall. Binary halts that all fire at once can cause the very panic they were meant to stop; graduated, de-synchronized friction is the safer design, even though the precise thresholds will be calibrated against real stress over time.

On insurance, the layer becomes first-class rather than an afterthought: mutualized reserve pools funded by a small skim on each instrument, specialized underwriters above them, and reinsurance absorbing the tail. The decisive novelty is that premiums price against live, observed correlation, and each insurer's own solvency is continuously verifiable. The last crisis's failed insurer did not fail simply because an insurer failed. It failed because it was opaque, undercapitalized, and perfectly correlated with what it insured, and no one could see it. Here, deterioration is visible before default, and capital binds on the insurance layer itself. This does not repeal correlation (a broad and fast enough tail event can still overwhelm any private layer), but it changes the speed and precision of the response.

And full-reserve money needs no backstop at all. There is no fractional-reserve leverage to unwind and no run on the unit of account to arrest. What is not automatically immune is the credit built atop the money: pools can face redemptions, and collateral can fire-sale. So the real question is liquidity provision to credit markets under stress, not deposit insurance.

The work is to recognize deterministic finality in payment-system law, so legal finality matches technical finality and a settled transaction is legally settled. To standardize correlation-aware margining with de-synchronized, graduated brakes: staggered, heterogeneous triggers, so the brakes do not become a monoculture that synchronizes the unwind they meant to dampen. To build a prudential regime for the insurance layer, with capital adequacy as a live constraint, not a quarterly attestation. And to resolve the backstop question with a private-first posture (over-collateralization, mutualized reserves, reinsurance, and pre-committed liquidity from large holders), leaving open whether a residual public backstop should reach the most systemically important credit infrastructure. What can be said with confidence is that where public intervention is ever needed, transparency makes it faster, smaller, and far better targeted than the blind bailouts of the past.

Reserve-yield distribution and the rent question

A fixed, non-yielding base unit throws off a large, concentrated income flow, and the question of who receives it is unavoidable. The base money pays no yield for being held; that is the firewall that keeps it safe. Under current statute, an issuer cannot pay interest to the holders of its money. So the yield earned on the reserves backing that money accrues, by default, to the issuer. This is a policy artifact. The flow goes to the issuer not by any natural law of money but because a statute put it there, and what statute creates, statute can redistribute. As the money supply grows to economy scale, the reserve income becomes a very large stream. Its default allocation is therefore a first-order distributional question, not a technicality.

The constructive point is that broad distribution is already the operating practice, not a hoped-for future. A leading regulated issuer already distributes the majority of its reserve income to its ecosystem (returned through distribution partners and usage-based rewards rather than as interest on the coin itself), and that share is rising, in pointed contrast to non-distributing models that keep the full float. The stakeholder-based network tokens emerging in this space are increasingly built so economic value flows by construction across validators, developers, and users. Value that statute bars from flowing to holders as interest can still flow to them as participation.

The author argues here against the commercial interest of the author's own industry, incumbents including issuers. The work is for policymakers to decide the contours of the interest prohibition deliberately rather than by inertia; to weigh whether more of the reserve income should be required to flow to holders and users; and to coordinate internationally, so differing rules on yield distribution do not become a vector for arbitrage. Treated as a question of public choice, the distribution of this income can be settled openly and well, and the architecture is strong enough to carry whatever answer that public process reaches.