Closed-Loop Currency

The problem

Stored value in 2017 was badly fragmented — gift cards, loyalty points, credit-card rewards, store credit, prepaid balances — and none of it composed. A customer with $40 of airline miles, $15 of Starbucks balance, $200 of a merchant gift card they didn't want, and a credit line could not combine those into a single purchase, nor could they trade unused balance with a second customer who actually wanted it. The merchant, the customer, and any lender financing the transaction were stuck in pairwise silos.

The mechanism

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 — customer, merchant, lender, co-customer, stored-value source — and the rules dictate, limit, or control how the transaction executes. Multiple stored-value sources combine within a single transaction ("use Customer A's unused gift card + Customer B's credit line + merchant loyalty points"), and when rule sets overlap or conflict, the system applies a hierarchy to resolve which rule governs. The architectural primitive is the rule set itself — programmable, composable, and enforceable across any number of parties.

What it proved

Value exchange is fundamentally a rule-based multi-party graph, not a bilateral ledger entry. If you can describe the constraints declaratively and resolve conflicts hierarchically, you can express any merchant/customer/lender/guarantor configuration as a single transaction. This is the conceptual ancestor of every "programmable money" pitch of the 2020s.

If it were built today

This family's 2026 reinterpretation is the strongest of the five, because the infrastructure finally exists. The 2017 rule-set primitive maps directly to smart contracts on stablecoin rails — USDC, PYUSD, RLUSD, EURC, and emerging state-issued CBDCs give you dollar-denominated programmable value with sub-second finality and deterministic execution. The $308B stablecoin market (Jan 2026) and $26T in 2025 on-chain volume is the substrate the patent was implicitly describing. A modern build: rule sets become smart-contract templates (ERC-4337 account abstraction, or Solana program accounts) that gate a transaction on multiple signatures from multiple stored-value sources; the hierarchy resolver becomes a policy engine (OPA-style) evaluated inside the contract; identifiers become verifiable credentials (W3C VC + DIDs) or zero-knowledge attestations of KYC status, creditworthiness, loyalty-tier, or merchant class. Critically, 2026 layers in agentic payments — the OpenAI/Stripe Agentic Commerce Protocol, x402 payment-required HTTP, Coinbase agentkit — so the "customers, merchants, lenders" of the original patent can all be AI agents composing a multi-party transaction autonomously. The hierarchy-conflict-resolution primitive is especially valuable here: when three agents each have policies that overlap on a transaction, you need a deterministic arbiter. The cleverness of the 2017 filing was seeing that the rules were the product; in 2026 that's the bet the entire stablecoin industry is making.