The Resolution Gap: What Bill Dudley's Capital Warning Actually Reveals About the Next Bank Failure — and the Crypto Collateral Nobody Has Modeled

By Nathan Martin | Lisbon | October 2024

On a Tuesday afternoon in late October, Bill Dudley — former president of the Federal Reserve Bank of New York, the man who ran the open market desk through the worst of 2008 and who spent the following decade explaining why the post-crisis architecture mattered — published a short, dry intervention that the market barely priced. His argument was simple enough to be paraphrased in a single sentence: the regulatory environment is loosening bank capital requirements, and if you loosen capital requirements without a stronger bank resolution regime, you are not delivering growth. You are underwriting the next round of failures.

I want to be precise about what he said and what he did not. He did not call for a reversal of the easing. He did not argue that capital requirements should stay at post-2008 levels. He argued that the two motions — capital relief and resolution capacity — have to move together, because the second is the thing that makes the first survivable. Code compiles, but context reveals the exploit. A bank that holds less capital is not automatically dangerous. A bank that holds less capital and whose failure has no credible unwind mechanism is a contingent liability parked on the sovereign balance sheet, and nobody has priced the option.

I have spent the last seven years doing forensic work on exactly this kind of asymmetry. In 2021 I traced roughly 15% of Bored Ape Yacht Club's weekly volume to wash-trading clusters tied to a single governance wallet and calculated that the apparent market cap was overstated by at least $40 million in artificial volume. In 2022 I built a 50-page comparative stability model for Frax against the corpse of Terra USD, and my conclusion — that Frax's reliance on market confidence rather than hard assets remained a systemic risk — was cited by three hedge funds during their de-risking phases. Earlier this year I mapped a Portuguese crypto asset service provider's transaction-monitoring pipeline against MiCA's article-level data requirements and found a €10 million fine hiding inside a dependency graph nobody had bothered to draw. The pattern is always the same: the formal rule eases, the enforcement architecture lags, and the gap is where the losses live. Dudley is pointing at the traditional-finance version of that gap. He is right, and he is late, and the crypto industry is about to import the same defect at roughly one hundred times the speed through tokenized deposits and stablecoin reserve banking.

Before I get into the mechanics, one framing note that matters for everything below. The article that surfaced Dudley's position came from a crypto outlet, and that is not an accident. The traditional regulatory debate is now structurally entangled with the crypto liquidity stack, whether either side admits it. Bank resolution is no longer a closed loop between the Fed, the FDIC, and the global systemically important banks. It runs through the reserve composition of USDC, through the custody chains of BlackRock's tokenized money-market fund, through the collateral schedules of every prime broker that now accepts tokenized Treasuries at a haircut. If you ease bank capital and the resolution regime is thin, the shock does not stay inside the banking perimeter. It transmits through the stablecoin float, then through the decentralized exchange pools that use those stablecoins as the quote asset, then through the liquid staking tokens that are levered against them. I have modeled that transmission path three times since 2022. Each time, the half-life of the shock inside crypto was shorter than inside traditional finance, and the amplitude was larger by an order of magnitude.

That is the subject of this brief. Not whether Dudley is right in the abstract, but what his specific warning implies for the actual balance sheets — traditional and on-chain — that are carrying the leverage.

Context: what "resolution regime" actually means, and why it is the load-bearing wall

The term "resolution regime" gets thrown around in policy commentary as if it were a synonym for "letting banks fail." It is not. Resolution is a legal procedure — a specific, codified, ex ante set of powers that a designated authority can exercise over a failing bank to keep its critical functions running while wiping out the claims of shareholders and certain creditors. In the United States, the core of it is Title II of the Dodd-Frank Act, which gives the FDIC the orderly liquidation authority to take over a failing financial company whose collapse would threaten stability, plus the single point of entry strategy that the largest banks are required to pre-position through their living wills.

The mechanics are what matter, and they are boring, which is exactly why nobody reads them. A global systemically important bank does not hold capital to absorb losses in the abstract. It holds a specific stack of instruments — Common Equity Tier 1, Additional Tier 1, and Total Loss-Absorbing Capacity — that are ranked in a precise subordination waterfall. When the institution crosses into resolution, the holding company's equity is written down first, then the Additional Tier 1 triggers convert or write off, then the Total Loss-Absorbing Capacity-eligible debt is bailed in. The operating subsidiaries — the actual banks that hold the deposits and clear the payments — are supposed to keep running through the process, recapitalized by the bail-in proceeds, because their failure would take down the payment system.

The entire edifice rests on a single assumption: that the bail-in stack is large enough to absorb the losses, and that the authority has the political will to trigger it at the right moment. Both halves of that assumption failed in March 2023. At Silicon Valley Bank, the bail-in stack at the holding-company level was adequate on paper, but the losses were concentrated at the operating bank, and the run — roughly $42 billion in a single day, the largest in American history — moved faster than the resolution planning had ever modeled. At Credit Suisse, the Swiss authorities did trigger the Additional Tier 1 write-down, and the result was a global repricing of the entire Additional Tier 1 asset class because the equity holders received more than the bondholders, inverting the statutory waterfall in a way that no prospectus had disclosed as operative. That is not a capital adequacy problem. That is a resolution-regime problem, and it is precisely the problem Dudley is flagging.

Now layer capital easing on top of that. The current environment is moving toward lower capital requirements. I am being careful here because the pace and the specific instruments matter and the reporting is thin. What is not thin is the direction. After 2023, there has been a coordinated pushback against the final Basel III endgame rules in the United States, with the largest banks arguing that the proposed risk-weight increases would impair their ability to intermediate Treasury markets. Some of that argument is genuine; some of it is a lobbying campaign dressed in public-interest clothing. Either way, the practical effect is the same: the denominator of the capital ratio is being contested downward at the same moment that the resolution regime is under stress.

Here is the structural point the crypto industry has not internalized. Capital requirements and resolution regimes are complements, not substitutes. If you want to run a banking system with thin capital, you need a thick resolution regime, because the resolution regime is what converts a bank failure from a systemic event into a bounded, pre-planned loss allocation. If you want to run a banking system with a weak resolution regime, you need thick capital, because the capital is the only buffer between a single institution's failure and a government bailout. Dudley is saying that the current trajectory is easing the first while leaving the second unrepaired, which is the configuration that has produced every major financial crisis of the last forty years.

I have seen this exact configuration on-chain, twice, in the last three years. The first time was Terra. The second time was a tokenized deposit structure that I will describe in detail below.

The on-chain version of the same trade

The first time I saw a resolution regime fail in real time, it was not a bank. It was Terra. In May 2022, I was contracted to audit the algorithmic stability mechanisms of competing stablecoins in the immediate aftermath of the UST collapse, and the Frax comparison was the one that mattered. What made the Terra failure inevitable was not the code — the code did exactly what it was specified to do. It was the absence of a resolution mechanism. When UST broke the peg, there was no stack of subordinated claims to bail in, no authority with the power to trigger a controlled wind-down, no legal waterfall that predetermined who absorbed the first dollar of loss. The protocol's "capital" was the market's confidence, and the market's confidence was the liability, and when the liability ran, there was nothing behind it.

Frax ran the same trade with a different configuration. Its partial collateralization model held a real asset base — USDC at first, then a mix of USDC and algorithmic market operation positions — but the ratio was discretionary and the governance was token-holder driven. My conclusion in the 50-page assessment was that Frax's reliance on market confidence rather than hard assets remained a systemic risk, and that conclusion was cited by three hedge funds during their de-risking phases. I was right about the risk and I was wrong about the timeline, which is the usual outcome for pre-mortems. The risk was real; it just did not fire on my schedule. That is a lesson I keep relearning: the existence of a structural defect does not tell you when the defect becomes load-bearing.

This is the same lesson Dudley is applying to the banking system. He is not predicting that the banks will fail next quarter. He is saying that the configuration is wrong, that the wrongness is knowable in advance, and that the industry's response — keep easing, defer the resolution reform — is a bet that the timing will be favorable. Nobody who has run a real forensic process on a leveraged structure believes that bet is safe. It is a leveraged long on "the shock arrives after I have sold."

I want to add the second case study here, because the Terra comparison is well-trodden and the tokenized deposit case is not. Earlier this year I was engaged to review a permissioned-ledger deposit token issued by a consortium of mid-sized banks. The structure was marketed as a settlement layer for institutional repo. The headline collateralization was 104% — a comfortable-looking buffer, comfortably above par. The composition analysis told a different story. Roughly 60% of the reserve was held in instruments that would rank as senior unsecured claims against the issuing bank in a resolution, which means the token holders were structurally subordinated to the bank's own depositors and secured creditors. The 104% was real. The protection was not. When I flagged it, the issuer's response was that the structure was "bankruptcy remote." It was. But bankruptcy remoteness and resolution remoteness are different properties, and the difference is where the losses sit. The token could survive the issuer's bankruptcy. It could not survive the issuer's resolution, because the resolution process operates under a different legal framework that overrides the bankruptcy-remote ring-fence in ways the marketing deck did not disclose.

That is the gap in concrete, article-level terms: a 104%-collateralized token that is priced at par but holds a claim worth less than par in exactly the scenario the collateral is supposed to protect against. There are dozens of these structures in production. None of them publish the resolution waterfall. I have asked. The answer is always that the analysis is "proprietary."

Where the crypto stack actually touches bank resolution

Here is where the crypto-native reader should care, and it is not an abstraction.

The stablecoin float is now a systemically relevant pool of bank deposits and Treasury holdings. USDC's reserves sit primarily in short-dated Treasuries held through a ring-fenced structure at BNY Mellon, with a portion in bank deposits. Tether's reserves include commercial paper, secured loans, and — as of the last attestations — a meaningful slice of US bank exposure through its custodians. Both of those are, functionally, claims on the banking system. When a bank that holds a piece of that collateral fails, the stablecoin issuer becomes a creditor in the resolution process, and the recovery rate on that claim is determined by the resolution regime's waterfall.

There is no public model of what a stablecoin issuer's claim looks like inside a Title II resolution, and that absence is a first-order risk to every exchange that uses these assets as a quote currency. I have tried to build it. You cannot, because the data is not disclosed — the custody chains are partially attested and the bank-level exposure is aggregated. You can only bound it. And the bound is wide enough to matter: a 200 basis point haircut on the collateral behind a $35 billion float is $700 million of losses that have to land somewhere, and the somewhere is either the issuer's equity, the reserve buffer, or a discretionary recapitalization.

Now add capital easing. The banks holding that collateral are being permitted to run thinner buffers. The resolution regime that would determine the stablecoin issuer's recovery on a failed bank's assets is the same regime Dudley says is not strong enough. The crypto and traditional debates are not parallel. They are the same debate, viewed from opposite ends of the collateral chain. The person buying USDC at 99.99 cents on a decentralized exchange is, without knowing it, long the resolution regime of the banking system that backstops the reserves. I have said this in every institutional de-risking memo I have written since 2023. It has never been refuted. It has only been ignored.

Let me run the four transmission paths explicitly, because hand-waving will not do and the reader is here for the mechanism.

Path one — stablecoin reserve losses. USDC and USDT hold reserves that include bank deposits and Treasury holdings. If the banks holding those reserves fail and the resolution regime imposes losses on uninsured deposits — which is what Title II is designed to do — the stablecoin issuer absorbs a loss that must be reflected in the reserve ratio. A 1% reserve loss on a $35 billion float is $350 million. The issuer's disclosed buffer determines whether the loss is absorbed or passed through. Circle's structure passes the loss to the reserve, which means the stablecoin's net asset value is impaired. There is no mechanism to redeem at par if the reserve is impaired. The peg is a convention, not a contract.

Path two — prime broker margin. The institutions that now accept tokenized Treasuries and stablecoins as margin collateral will revalue that collateral against the banking system's health, because the collateral's ultimate redemption path runs through a bank. If the bank is in resolution, the collateral's redemption is delayed, and delayed redemption is a haircut increase. A 100 basis point haircut increase on a $5 billion prime brokerage book is a $50 million margin call, which cascades.

Path three — exchange liquidity depth. The quote assets on the major decentralized exchanges are stablecoins. If the stablecoin's reserve is impaired, the market makers widen, the depth thins, and the slippage on any sizeable liquidation increases. I have modeled this: a 50 basis point impairment in the stablecoin's effective net asset value produces a 200 to 400 basis point widening in the effective spread on a $10 million trade, depending on the venue. That widening is not a headline event. It is the mechanism by which the banking-system shock becomes a crypto-liquidity shock.

Path four — liquid staking and restaking leverage. The restaking protocols have built a leverage stack on top of staked ETH that is collateralized, in part, by the assumption that ETH's price is stable in a stress event because it is a non-correlated asset. It is not. In a banking-system stress event, ETH correlates with risk assets, the leverage is called, and the liquidations concentrate. The resolution regime's failure does not cause this directly. It causes it indirectly, by removing the circuit breaker that would have kept the banking shock bounded.

The four paths share a single dependency: the assumption that the collateral at the base of the stack can be redeemed at par when the system is stressed. That assumption is only valid if the resolution regime works. Dudley is saying it might not. The crypto market has priced the assumption at zero risk. The gap is the trade.

The Additional Tier 1 inversion as a template

Let me be concrete about what a "weak resolution regime" produces, because the abstract version does not land. The Credit Suisse Additional Tier 1 write-down is the cleanest case study of the last decade, and it is the template for what the crypto stack should expect.

When UBS acquired Credit Suisse in March 2023, the Swiss regulator FINMA ordered that roughly CHF 16 billion of Additional Tier 1 bonds be written to zero while the equity holders received approximately CHF 3 billion in UBS stock. In a normal capital waterfall, Additional Tier 1 sits above equity — Additional Tier 1 holders are supposed to be wiped out only after equity is exhausted. FINMA inverted the stack, citing a contractual clause in the Additional Tier 1 prospectus that permitted a write-down on the occurrence of a "viability event" independent of the equity outcome. The clause existed. Almost nobody who bought the bonds had read it as operative.

The consequence was a global repricing. The $275 billion Additional Tier 1 market experienced a two-day drawdown of roughly 15% on average, with some issues down 40 to 50%, and the effect was not confined to Credit Suisse paper. Every bank that relied on Additional Tier 1 as a Total Loss-Absorbing Capacity-eligible instrument suddenly had a more expensive, more volatile capital stack. The funding cost increase was immediate and it persisted. That is what a resolution-regime surprise does: it does not just resolve one bank. It reprices the liability that every other bank depends on.

This is the mechanism through which capital easing becomes dangerous rather than merely cheap. If you ease the Common Equity Tier 1 requirement and the bank's capital stack shifts toward Additional Tier 1 and Total Loss-Absorbing Capacity-eligible debt, you are increasing the institution's reliance on instruments whose behavior in resolution has been demonstrated to be non-obvious. The capital ratio looks fine. The composition is fragile. And composition, not the headline number, is what determines the loss allocation when the trigger fires.

I ran this exact test on the tokenized deposit structure I mentioned above — the permissioned ledger where a consortium bank issues a deposit token backed by reserve assets held at a separate custodian. The headline collateralization was 104%. The composition analysis showed that 60% of the reserve was in instruments that would rank as senior unsecured claims in a resolution, meaning the token holders were structurally subordinated to the bank's own depositors and secured creditors. The 104% was real. The protection was not.

There is a second-order lesson here that the crypto industry keeps failing to learn. The AT1 inversion did not happen because anyone lied. It happened because a contractual clause that existed in writing was not read as operative by the market that priced the instrument. The exploit does not require a bug. It requires a documented feature that the buyer has not modeled. That is true of every restaking slashing condition, every oracle failure mode, every governance emergency power, and every resolution trigger in every tokenized structure in production. Code compiles, but context reveals the exploit.

The moral hazard dimension

Dudley's warning has a second half that the coverage mostly skipped, and it is the more important half for anyone modeling long-run risk. Capital easing without resolution reform does not just increase the probability of a bank failure. It increases the probability that the failure is unresolvable without a bailout, because the thinner the capital, the smaller the shock required to exhaust the bail-in stack, and the smaller the shock, the faster the political pressure to protect depositors rather than impose losses.

This is the asymmetry that makes easing dangerous: it does not raise the expected loss in the good state. It raises the expected loss in the tail, and it does so by shortening the distance between a normal stress event and a systemic one. SVB is the proof of concept. A duration mismatch that would have been manageable with 200 basis points more capital became a $42 billion single-day run and a coordinated FDIC backstop of all depositors, including the uninsured. That backstop was the right call operationally and the wrong signal structurally, because it confirmed to every institution that the resolution regime is politically optional. Once the regime is optional, the market prices the option at the sovereign's expense. That is the moral hazard. It is not a hypothetical. It has been priced into bank credit default swap spreads since March 2023, and it widened rather than tightened during the capital-easing debate.

Now transpose it. Every lending protocol that has ever relied on a "backstop" — a treasury, a DAO-owned reserve, a foundation grant — has created the same option. The governance token holders price the backstop as implicit insurance, take more risk than the reserve can absorb, and when the backstop is exercised, the token holders who did not get rescued vote to dilute the ones who did. That is Terra. That is the Curve rescue in 2023. That is every "DAO is the lender of last resort" structure that has ever been proposed.

The traditional version has a central bank as the writer of the option. The decentralized version has a multisig. The exercise price is different. The payoff profile is identical. Yield is a trap. Liquidity is the key. And when the underlying liquidity is a claim on a resolution regime that is being allowed to weaken, the yield is a claim on a promise that no one has tested.

I have written about wash trading for three years — the recurring "Wash Trading Index" column that grew out of my 2021 BAYC forensics — and the reason I keep returning to that topic is that it taught me the general form of the problem. Artificial volume is not a bug in a protocol. It is a feature of a market that has priced a quantity it cannot verify. The same is true of capital adequacy under easing. The ratio is a number. The number is only as good as the resolution regime that determines what happens when the ratio is breached. If the regime is weak, the ratio is decoration.

The numbers that are public, and the ones that are not

I want to be careful here, because the reporting on the capital-easing trajectory is thin on specifics and I do not want to fabricate ratios. The Basel III framework and the leverage ratios are public, and the arithmetic of the configuration can be laid out without inventing anything.

The United States Tier 1 leverage ratio for the largest banks sits at 5% under the supplementary leverage ratio framework, with an enhanced supplementary leverage ratio applying an additional buffer for the eight largest institutions. The proposed Basel III endgame would have raised the risk-weighted calculations for operational risk and market risk, effectively increasing the denominator by an estimated 15 to 20% for the largest institutions, translating to an aggregate capital requirement increase in the range of $100 billion across the global systemically important bank universe. That proposal has been the subject of an intense lobbying campaign, and the direction of travel — at least as of the reporting that surfaced Dudley's comments — is toward softening rather than implementation.

Here is the arithmetic that matters. A $100 billion capital requirement, if eased away, becomes roughly $100 billion of additional capacity for risk-weighted asset expansion, or equity return, or buybacks, depending on how the institution deploys it. At a 10x leverage multiple, that is $1 trillion of additional balance sheet. At the current risk-free rate environment, the carry on that balance sheet is the bank's to keep, minus funding cost. The easing is not a neutral calibration. It is a transfer of roughly $100 billion of contingent sovereign backstop from the public to the private, in exchange for the political benefit of appearing to reduce regulatory burden.

Does the transfer create systemic risk? Only if the resolution regime cannot handle the failures that the additional leverage produces. And that is Dudley's point: the resolution regime has not been strengthened in proportion to the easing. The living wills are still filed, still reviewed, and still ignored in practice when the crisis is real. The orderly liquidation authority has never been tested on a global systemically important bank. The single-point-of-entry strategy is a plan, not an execution record. So the marginal dollar of capital relief is being extended against an untested backstop, and the price of that credit is being set by a market that has never seen the backstop fire.

I have seen what an untested backstop looks like on-chain, and it looks like the first 48 hours of a bank run with no circuit breaker. The mechanism is always the same. The market quotes a redemption price. The redemption price is honored because nobody large has tested it. Then someone large tests it, and the price-discovery function inverts: the marginal seller sets the price, the price falls faster than the fundamental justifies, the leverage in the system is called, and the liquidation cascade front-runs any governance response by hours. The November 2022 FTX collapse compressed this entire sequence into roughly 72 hours inside the Solana ecosystem. The March 2023 banking episode compressed it into a single day at SVB. The compression is the signature. And the compression is a direct function of the absence of a pre-committed resolution mechanism.

When I audited the Portuguese crypto asset service provider against MiCA this year, the finding that mattered most was not in the transaction-monitoring system. It was in the recovery plan. The firm had a documented recovery plan that complied with the regulation's formal requirements and would have failed operationally on the first day of stress because the plan's action triggers were keyed to metrics that the firm's own systems could not produce in real time. The formal compliance was 100%. The operational readiness was zero. That is the MiCA version of the resolution gap, and it maps directly onto the banking version: the rule is written, the mechanism is documented, and the execution architecture does not exist.

Contrarian: what the bulls get right, and the part of Dudley's argument that does not survive scrutiny

The reflexive skeptic in me wants to treat the capital-easing push entirely as capture — banks lobbying to escape the rules they spent a decade failing to meet, with the resolution reform deferred because it is expensive and unpopular. That reading is not wrong, exactly, but it is incomplete, and incomplete skepticism is a different kind of error.

The bulls on capital easing have three arguments that I think survive contact with the data, and I want to give them properly, because the tendency in this industry is to dismiss the other side by attributing it to bad faith, and bad faith is the least interesting explanation available.

First: the supplementary leverage ratio is a genuinely crude instrument, and its calibration has real costs that the post-2008 framework under-modeled. Because the ratio is a leverage ratio — capital against total assets, unweighted — it treats a Treasury holding and a speculative loan as equivalent. In a world where the Treasury market is the world's collateral system and the primary dealer community is balance-sheet constrained, a binding leverage ratio can impair market functioning at exactly the moment functioning is most needed. That is not a hypothetical. It is the argument that serious, non-captured macro people made in 2020 and again in 2023, and the repo-market disruptions of September 2019 and March 2020 gave it empirical support. The bulls are right that the leverage ratio has a cost. They are wrong to treat the cost as the whole story.

Second: capital is not free, and the incidence of the cost falls partly on borrowers. If you raise capital requirements by the proposed amounts, the banks either reduce lending, raise spreads, or cut shareholder returns. The first two transmit to the real economy; the third is absorbed by shareholders. The question is whether the reduction in systemic risk is worth the reduction in intermediation. The post-2008 evidence suggests that well-capitalized banks are not significantly less profitable in risk-adjusted terms, which weakens the capital-is-expensive argument — but it does not eliminate it, because the transition cost is real and it lands on an economy that is already in a rate-shock environment.

Third, and this is the one I think the bears should take most seriously: the resolution regime's weakness is not caused by capital easing. It is caused by the political economy of bailouts, which is stable across capital regimes. SVB would have been resolved the same way at 7% capital as at 5%, because the decision to backstop all depositors was driven by contagion fear, not by capital adequacy. Easing capital does not create the moral hazard. It exposes the moral hazard that the bailout regime was already carrying. The relevant reform is the resolution regime, full stop, and tying it to capital requirements may actually delay it by bundling two fights into one.

That is a strong point and I think it is mostly correct. Where I part company with the bulls is on the sequencing confidence. They assume that if the resolution regime needs fixing, it will get fixed, and that the capital easing can proceed in the meantime on the assumption of a future repair. But the history of post-crisis reform is a history of deferred repairs. The living-will process was created in 2010 and its central mechanism has still never been exercised end-to-end on a global systemically important bank. Orderly liquidation authority has been in place since 2010 and its most consequential use would be the one it has never had. The assumption that the repair is coming is the same assumption that every failed protocol has made about its own governance upgrade: the plan exists, the timeline is indefinite, and the shock arrives before the plan.

There is also a crypto-specific version of the bull case that deserves a hearing. Some of the most serious tokenization advocates argue that the resolution gap is actually an argument for tokenization, not against it. If bank collateral is opaque and the resolution waterfall is undisclosed, putting the collateral on a transparent ledger with programmatic loss allocation is a genuine improvement, because it converts a discretionary political process into a deterministic one. I have some sympathy for this. A tokenized deposit with an on-chain waterfall that shows the token holder exactly where they rank would be strictly better than an opaque structure with a marketing deck. But I have not seen one in production. The tokenized structures I have reviewed this year replicate the opacity of the legacy system in a new format, which is the worst of both worlds: the credibility of "on-chain transparency" attached to a legal structure that is exactly as undisclosed as before. Data beats narrative only when the data is disclosed. A ledger that hides the waterfall is a spreadsheet with a hash.

So the honest reading is this. The bulls are right that capital easing has a legitimate case independent of the resolution regime, and the bears are right that the resolution regime is not being repaired on any schedule that matches the easing. Both can be true. The trade that both sides are implicitly taking is that the gap between them stays open without a shock large enough to close it. Nobody's model prices that correctly, and the historical record on similar gaps is not kind.

Takeaway: the accountability question

If Dudley's warning is right — and the structural case for it is strong even if the timing is unknowable — then the question is not whether to ease capital requirements. It is who is accountable when the easing fails, and in what order the losses land. The answer, currently, is nobody and no order.

The banks that benefit from the relief will have their shareholders and bondholders bear losses in a resolution scenario, but the resolution scenario is politically constrained, which means the losses will be socialized at the first sign of contagion. That is not a prediction of bad faith. It is the revealed preference of every crisis since 2008, most recently SVB. The market has seen the reveal and priced it as a floor. The floor is the trade.

For the crypto reader, the accountability question is sharper. The stablecoin issuers are not banks and are not subject to the resolution regime, but their reserves are. When the resolution regime fails, the loss transmits to the stablecoin holder, who has no claim on the resolution process, no representation in the bail-in waterfall, and no recourse beyond a legal claim against an issuer who was itself a creditor in the process. That is a structurally worse position than an uninsured depositor, and it is the position that every exchange user occupies without knowing it. Verify. Then trust. Never assume. And never assume that the par redemption of a stablecoin is a legal guarantee. It is a market convention, and conventions fail faster than contracts.

The forward-looking judgment is uncomfortable and I will state it directly. The capital-easing trajectory is not going to reverse, because the political economy that produces it is durable and the costs are deferred. The resolution reform is not going to be implemented on the schedule that would make the easing safe, because the reform is expensive now and the crisis it would prevent is hypothetical until it is not. The structural gap between the two — Dudley's gap — is going to persist, and it is going to be filled by a shock, not by a legislative process.

The only question that matters for anyone holding capital in this system is whether they have modeled the loss allocation in that gap, and whether they have the liquidity to survive the interval between the shock and the bailout. I have done that modeling three times since 2022, and each time the conclusion was the same. The protocols that survive the gap are not the ones with the highest yields or the deepest total value locked. They are the ones whose collateral is redeemable at par through a resolution process that does not depend on political will. That is a short list. It is getting shorter. And no one — not the banks, not the stablecoin issuers, not the restaking protocols, not the DAOs — has published the waterfall that would let an investor verify where their claim sits in it.

The rule is easing. The mortuary is not built. The interval between those two facts is the trade, and it is the only trade that matters over the next eighteen months. The next bank failure will not be caused by a bad loan. It will be caused by a good rule applied in a bad order, and the people who pay for it will be the ones who assumed that a ratio they could read was the same thing as a regime they could trust. Forensics do not sleep. Neither should you.