The rules are the product.

The Family 5 patent filed in January 2017 — Closed Loop Currency Exchange, WO/2018/140261 — opens with a sentence that does most of the work: a closed-loop value exchange defined by one or more rule sets, where each rule set associates two or more unique identifiers with a transaction. Everything after that is mechanism: customer, merchant, lender, co-customer, stored-value source; multiple stored-value sources compose within a single transaction; a hierarchy resolves overlapping rules. The structure is almost embarrassingly simple. The primitive is the rule set. The product is the rule language.

That was not the dominant framing of value exchange in 2017. Gift cards were closed-loop silos. Credit card networks were bilateral ledgers. Loyalty points were an accounting liability trapped inside one brand. Stablecoins existed — USDT had been around since 2014 — but nothing on top of them looked remotely like composable programmable value. The filing was a claim about architecture, not a product, and it was made against a landscape that did not yet have the rails to run it on.

Nine years later, the rails showed up. This is what they look like, and what the 2017 filing actually claimed they would become.

What the rule-set primitive actually specifies

Strip the patent down to its atoms and it defines four things. First, an identifier registry: parties to a transaction are identified by codes that don't have to be universal — they can be domain-scoped, anonymized, or cryptographically derived. Second, a stored-value adapter: any source of value (gift card, credit line, points balance, bank account) can participate in a transaction through a common interface. Third, a rule-set definition: a set of constraints binding identifiers to permitted behaviors, enforceable at transaction time. Fourth — and this is the part that ages best — a hierarchy resolver: when multiple rule sets apply to the same transaction and they conflict, a deterministic process picks which rule governs.

The hierarchy resolver is the quiet piece. In 2017 it mattered because a single transaction might blend merchant rules, loyalty rules, and lender rules that disagreed about, say, whether a stored-value source could cover a particular SKU class. In 2026 it matters for a completely different reason: when three autonomous agents are each committing value on behalf of their principals, you need a deterministic arbiter, not a negotiation. The 2017 text didn't say "agent," because agents were not yet transaction parties. The structure it specified does not care.

What 2022-2026 shipped

Stablecoins at scale.USDC, USDT, PYUSD, RLUSD, and EURC cleared roughly $26 trillion of on-chain volume in 2025. Stablecoin supply crossed $308 billion in January 2026, and the market has diversified away from a single issuer monopoly toward a handful of regulated operators. This is the settlement layer the patent's atomic-transaction orchestrator needed — dollar-denominated, deterministic, sub-second-final, and programmable. In 2017 the closest thing was a bilateral ledger write. In 2026 it is a smart-contract call.

Account abstraction and programmable wallets.ERC-4337 account abstraction, Solana program accounts, and equivalent primitives on other chains let a single wallet be governed by arbitrary code — multi-signature, session keys, spending limits, source-of-funds rules, time locks. A 2026 wallet is itself a rule-set engine. The patent's "rule set binding identifiers to a transaction" maps directly onto a smart-contract-governed wallet; what the 2017 filing called a rule set, Ethereum calls a contract module.

Verifiable credentials and decentralized identifiers. W3C VCs and DIDs provide the identifier layer the patent specified but could only describe in the abstract. A VC can attest to KYC status, accreditation, creditworthiness, loyalty tier, or merchant class without exposing the underlying personal data. A DID is the stable handle by which a party — human, corporate, or machine — participates in a transaction. The identifier registry in Section 1 of the rule set is now a cryptographic object, not a database row.

Agentic payments and the x402 / ACP layer.The OpenAI/Stripe Agentic Commerce Protocol, the x402 payment-required HTTP standard, Coinbase's agentkit, and the emerging MCP commerce tools let software agents be first-class transaction parties. This is the change that makes the hierarchy resolver indispensable. When a procurement agent, a supplier agent, and a financing agent each have policies committed to the same transaction, the settlement can't wait for humans to arbitrate — the rule set has to decide, in real time, reversibly, auditably. The 2017 filing specified exactly that behavior. Nobody was building for it at the time.

What 2026 still doesn't have

Given all of the above, what's missing? Three things, in ascending order of consequence.

A standard rule language.Stablecoins are denominated in dollars; account-abstraction contracts are written in Solidity, Cairo, or Anchor; policy engines are OPA, Cedar, or in-house. There is no cross-platform rule language that expresses "insurer-approved amount before HSA before credit, patient copay last, rebates require supplier certification, settlement atomic or reverted." Every implementation rebuilds the same rule vocabulary from scratch. The market that crystallizes this standard captures the compiler layer of programmable money.

A cross-rail compositor.A healthcare checkout wants to combine HSA dollars, insurer-approved benefit dollars, and out-of-pocket stablecoin settlement in a single atomic transaction. The HSA sits on ACH rails, the insurer settles through CHIPS or real-time rails, the stablecoin sits on-chain. The patent's transaction orchestrator was rail-agnostic; 2026 production systems are not. The plumbing that makes five rails behave as one transaction is not yet standardized.

A market for the instruments themselves. Composable programmable value requires not just settlement rails but a venue where the resulting rule-governed instruments trade, price, and clear. This is the most under-built of the three. It is also the specific problem that the American Innovation Economy framework I've been developing addresses directly, through Innovation Credits.

Innovation Credits as an instantiation

Innovation Credits are a zero-coupon sovereign instrument class denominated in U.S. dollars, backed by verified and insured domestic innovation yield, that foreign importers purchase to access the American market. Mechanically, they do what a tariff does — transfer value from foreign buyer into the U.S. fiscal system — but they produce a tradeable asset on the Treasury's balance sheet instead of disappearing into a revenue line. Held to maturity they pay face value; sold on secondary markets they trade at discount; used as collateral they extend into credit and liquidity products.

The reason Innovation Credits matter for this essay: every transaction that flows through the Innovation Credits system is, structurally, a Family 5 closed-loop exchange. Multiple parties (importer, Treasury, retailer, consumer) with multiple stored-value sources (fiat, stablecoin, credit, point-of-sale residual) are bound by a rule set (duties, compliance constraints, redemption rights) with a hierarchy resolver (which rule governs when a retailer's loyalty terms conflict with the credit's face-value redemption terms). The instrument is the packaging. The rule set underneath is the primitive.

The downstream consumer-redemption pathway — residual credit value transferring to consumers at point of sale, attached at the SKU level via GS1 Application Identifier 8112 at retailers like CVS — is the closed-loop-on-open-loop move spelled out in the patent: a closed-loop program delivers the brand-and-policy experience, the open-loop settlement layer delivers the economics. In 2017 that was an abstraction. In 2026 it's wired through BNY Mellon/Pershing clearing and production retail POS systems.

What the commercialization question actually is

The question stops being "can programmable settlement work" and starts being "whose rule language does it speak." The settlement layer is commoditized. The wallet layer is competitive but converging. The differentiator is the policy-and-composition surface — the rule-set vocabulary itself, the hierarchy-resolution semantics, the agent-callable interfaces, and the audit/reversibility guarantees bolted on top.

This is the part of the stack the 2017 filing was actually about. Nine years of infrastructure build-out has produced the rails; what sits on top is still contested, and the winner captures an enormous amount of the economics. For the portfolio specifically, Family 5 sits exactly where that contest is happening. The patent does not claim the stablecoin, the wallet, or the agent. It claims the rule set that binds them.

What I'm building now, in one paragraph

Hardware-trust credentials provisioned directly into the Apple Secure Element. Software-defined point of sale on commercial-off-the-shelf devices, MPOC certified. Cross-brand loyalty exchange that composes points across brand portfolios. Treasury-as-a-service that manages float and inter-company liquidity for multi-brand groups. Open-loop transit acceptance that quietly does the largest unbanked-inclusion work of a generation. Stablecoin onramps and agent-callable payment tools on top of all of it. Seven surfaces. One primitive underneath. The primitive is the rule set. The product is the rule language. The portfolio that describes it is available.