The classifier function inside 31 CFR §1523.1(c) is four lines long. It decides whether a firm is a Payment Stablecoin Issuer or a Digital Asset Service Provider. It decides who posts capital and who merely files suspicious activity reports.
// 31 CFR §1523.1(c) — simplified
function classify(entity) returns (Role) {
if (entity.obligatedToRedeemAtPar()
&& entity.issuesToPublic()
&& entity.holdsStableValue()) {
return PPSI; // licensed entity — heaviest burden
}
if (entity.transfersOnly() || entity.custodiesOnly()) {
return DASP; // AML/sanctions gatekeeping — lighter
}
return UNCLASSIFIED; // the branch nobody is writing about
}Four lines. No upgrade path, no grace period on the classification itself. Treasury published this as a Notice of Proposed Rulemaking, and the comment window closes 19 October 2026. The obligations begin 18 January 2027 whether or not the comment window produced anything coherent.
That asymmetry — a fixed compliance clock bolted to an open definitional clock — is the thing I keep circling. Not the politics. The state machine.
Context
The GENIUS Act established the legislative frame for payment stablecoins in the United States. The NPRM is the implementation layer, and implementation layers are where architecture actually gets decided. Legislators write intent. Regulators write control flow.
What Treasury built here is a two-tier regulatory topology. The upper tier is the issuer: a licensed entity required to hold reserves in short-dated Treasuries and similar high-liquidity instruments, obligated to redeem at par, and prohibited from paying interest or yield to holders. The lower tier is the DASP: an entity that moves, custodies, or distributes stablecoins but does not carry the par-redemption obligation, and is therefore gated primarily on AML and sanctions compliance rather than capital.
Code does not lie, but it does hide. Read the function above and it looks like a clean partition. Read the comment letters and it is not a partition at all. The Bank Policy Institute, filing jointly with the American Bankers Association, the Consumer Bankers Association, the Financial Services Forum, and The Clearing House, is arguing that the interest and yield prohibition should extend to the DASP tier as well. Their framing is "Safety Across the Ecosystem." The operational content of that framing is: make the distribution layer carry the same restrictions as the issuance layer.
If that lands, the light tier stops being light.
The timeline Treasury attached to this is unusually rigid for a rule that is not yet final. Conditional OCC approvals went to Ripple, BitGo, Fidelity, and Paxos in December 2025. Circle received the only final approval — a national trust bank charter under the name First National Digital Currency Bank — in July 2026. The comment period on the NPRM runs to October 2026. GENIUS Act obligations and restrictions on foreign stablecoin issuers activate 18 January 2027. Full restriction on stablecoins issued by non-licensed entities activates 18 July 2028.
Four dates. Two of them are definitional, two of them are enforcement. They do not move together.
Core Analysis
The economic substance test is the load-bearing member of this structure, and it is worth being precise about why. Regulatory frameworks usually classify by activity list: you did X, therefore you are Y. Activity lists are stable, auditable, and trivially gameable, because the entity being classified controls the description of its own activity. Economic substance tests classify by function instead. The question is not what you say you do. The question is whether you stand behind a par-redemption obligation.
That is a better design, and it has a specific consequence that I do not think has been adequately priced.
Par-redemption-at-will is a hard invariant. It requires a reserve that is, under stress, actually liquid at par. I spent three weeks in 2021 reverse-engineering the cross-chain signature verification path in the Poly Network bridge, mapping the byte-level discrepancy in the access control list that let an attacker mutate state that should have been immutable. The lesson I took from that autopsy was not "multisig bad." It was that a structural flaw is invisible in the specification and obvious in the execution path. The spec said the bridge was guarded. The runtime said otherwise.
The same lens applies here. An entity that issues a stablecoin backed by a fractional reserve, or an algorithmic peg, or a partially collateralized basket, cannot credibly commit to par redemption across all states. Not because the team is dishonest — because the invariant does not hold in every branch. Under §1523.1(c), that entity does not get to be a PPSI. It also does not cleanly fit DASP, because it is issuing, not distributing. It falls into the third branch.
UNCLASSIFIED is not a safe harbor. It is an unbounded loop.
Infinite loops are the only honest voids. They do not resolve; they consume. An issuer that cannot satisfy the par-redemption predicate and cannot satisfy the transfer-only predicate has no compliant configuration available to it under this rule set. That is not an oversight. That is a filter with a zero-width passband for an entire class of designs.
Now look at the license layer, because that is where the filter converts into market structure. Four conditional approvals and one final approval is not a pipeline. It is a throttle. The OCC has effectively created a scarce resource, and the scarcity is not accidental — it is the mechanism. Root keys are merely trust in hexadecimal form. A national trust bank charter is the same thing with a different encoding: it is a transferable, revocable, concentrated permission to be the entity that stands behind par redemption.
Circle being the sole final approval is not a marketing claim. It is a measured asymmetry in a state variable that competitors cannot shortcut with engineering. No amount of contract optimization closes a charter gap. In 2024 I spent a quarter refactoring a SNARK verifier's constraint system — redundant modular arithmetic in the verifier contract was inflating gas by roughly 40 percent, and we cut it. That work mattered because cryptographic soundness has to survive economic viability or it does not ship. The stablecoin charter layer inverts that relationship. The soundness here is administrative, and it does not care about your gas costs.
The distribution layer tells the opposite story, and this is where the architecture gets genuinely interesting. Coinbase, through Stablecore and Moov, is reporting coverage of more than 3,000 community banks and credit unions on one side and more than 4,000 distribution endpoints on the other. That is not a marketing number either. That is an integration surface, and integration surfaces compound.
Here is the mechanical reason the DASP tier is worth more than its regulatory burden suggests. The prohibition on paying interest or yield to stablecoin holders applies to issuers. It does not, in the current text, apply to distributors. That gap is a value-capture channel. An issuer earns reserve interest and cannot share it. A distributor that is not an issuer can build a rewards mechanic that looks like yield without being yield — fee rebates, loyalty accrual, promotional credits. The economics are identical from the holder's perspective. The regulator's classifier does not see it, because the classifier asks about redemption obligations, not about marketing programs.
Velocity exposes what static analysis cannot see. The rule text is a snapshot. The distribution layer is a flow. And the flow is going to find the seam.
This is precisely why BPI wants the prohibition extended downward. Their comment is not an AML argument dressed in safety language — it is a competitive argument, and it is a competent one. If the prohibition reaches DASPs, the seam closes. Distributors lose the ability to differentiate on economics, revert to pure fee models, and converge on the bank compliance posture. The moat gets rebuilt around institutions that already have compliance departments.
The vertical integration signal confirms that sophisticated players have already read the seam. Stripe acquiring Bridge. Mastercard acquiring BVNK. Both moves put the infrastructure layer under a single balance sheet, which means both moves are bets that the next three years of regulatory text will make the issuance tier expensive and the distribution tier decisive. If you own both, you do not need to predict which tier wins.
Note the structural detail that most coverage skips: the NPRM does not make PPSI and DASP status mutually exclusive. A single entity can hold both roles. The architecture explicitly permits a firm to be both the obligor and the distributor. That is not a loophole; it is a deliberate design choice that rewards full-stack consolidation and punishes pure-play positioning. The regulatory topology and the market topology are converging on the same shape.
Now the clock. This is where the design has a defect, and I want to name it clearly.
Obligations activate 18 January 2027. Full restriction on non-licensed issuers activates 18 July 2028. Those dates are statutory. They do not slide if the agencies miss their rule-writing deadlines. Six agencies were assigned a July 2026 rulemaking deadline; not all met it. So the plausible failure mode is not delay. It is the opposite: obligations arrive before definitions do.
Security is a process, not a product, and a process that outruns its own specification produces exactly one outcome. Firms configure to the best available interpretation, and then get reclassified retroactively when the final text lands. The comment window closing in October 2026 leaves roughly three months before activation. Three months is not enough time for a distribution business to rebuild its compliance stack, particularly if the yield prohibition has been extended and a rewards mechanic has to be unwound.
Contrarian Angle
The consensus read on this rule set is "regulatory clarity, structurally bullish for compliant issuers." I think that reading is correct and incomplete in a way that matters.
The most consequential provision in the entire framework is not the issuer definition. It is the 2028 restriction on non-licensed issuers. Read it against the 2027 restriction on foreign issuers. Two dates, eighteen months apart, both pointing at the same class of entity. The first requires foreign issuers to comply with lawful orders. The second prohibits distribution of stablecoins issued by anyone without a license.
A $100B-plus offshore dollar stablecoin cannot obtain a US trust bank charter without submitting to a reserve, audit, and redemption regime it has spent its entire existence avoiding. So the path is not compliance. The path is exclusion from the largest regulated distribution surface in the world.
That is a structural eviction schedule, written in calendar form. It is the largest unpriced market event embedded in this document, and it is discussed almost nowhere, because it is not stated as a policy goal — it is stated as a compliance date.
Code does not lie, but it does hide. So do effective dates.
The second blind spot is the security framing itself. The interest prohibition is usually described as consumer protection or as a guard against stablecoins being marketed as investments. Both are true. Neither is the interesting part. The functionally important effect of the prohibition is that it removes the "expectation of profits" element from the classification analysis entirely, which is what lets payment stablecoins sit outside the investment contract frame. The prohibition is not a restriction bolted onto the product. It is the mechanism that makes the product's regulatory identity coherent. Remove it and you do not get a freer stablecoin. You get a security.
Which means the fight over extending it to DASPs is not a fight about safety at all. It is a fight about who gets to run a rewards program.
Takeaway
The rule text is a snapshot; the distribution layer is a flow. Within eighteen months, watch three variables and ignore the rest: whether the prohibition reaches DASPs, whether a second OCC final approval appears, and whether the six agencies close the rule gap before January 2027.
If the prohibition extends and the approvals stay concentrated, the US stablecoin market becomes a licensed oligopoly distributing through bank rails — and the offshore dollar stablecoin problem resolves itself by eviction rather than by competition. If the prohibition stays scoped to issuers, the distribution tier becomes the most valuable real estate in payments, and the firms that bought their way into it will look prescient.
Either way, the four-line classifier at §1523.1(c) is doing more work than any statute in this space has done before. The question is not whether it is correct. The question is whether anyone has tested the UNCLASSIFIED branch before a few hundred million dollars of supply lands in it.