A token names a denomination and a contract. It does not say what stands behind it — that is the issuer's reserve, and a bank may back instruments with cash, with bonds, or with a stablecoin it holds. The choice changes the economics and the regulation, and it changes one design decision inside the token.
The constraint first. A token's denomination is inside the preimage of the issuer's signature, so it cannot change. An instrument therefore cannot "accrue" by growing: an interest-bearing token is not a rebasing balance, it is a fixed face value plus interest that accrues beside it and is paid as new instruments. Everything below follows from that.
| Reserve | What the token is a claim on | What it does to the design |
|---|---|---|
| Cash at the issuer | A deposit, or central-bank money where the bank has it | Stable face value, no yield. The simplest case and the one the sandbox models. |
| Bonds | A pool of fixed-income assets; coupons are the funding source | Face value fixed, yield accrues to holders. Requires maturities, duration and a policy for what happens if the pool is sold at a loss. |
| A stablecoin | Whatever backs that stablecoin | One more layer: the token is a claim on a claim. Fine if the issuer accepts the counterparty risk knowingly, and it should be stated as a two-step chain, not hidden as "cash". |
Three models, in increasing complexity. Only the first two are compatible with a fixed signed denomination without re-issuance.
1 holder of record coupons accrue in the record; paid at redemption as face + accrued 2 periodic distribution interest is issued as new tokens on a schedule (monthly, quarterly) 3 rebasing the balance grows in place — requires re-signing, so it is not this design
Model 2 is what a bank usually wants because holders can see it: on the first of the month, each holder with a bond-backed instrument is issued small tokens equal to its share of the coupon. It costs instrument slots, so it suits float-sized balances and not thousands of tiny accounts.
A token's denomination is a unit as well as an amount. Choose the unit deliberately and three very different instruments come out of the same machinery:
| Anchor | Unit on the token | Yield | What moves |
|---|---|---|---|
| Stablecoin | Currency, matching the stablecoin's unit | yes — accrued in the record, paid at cash-out | Nothing, as long as the stablecoin holds its peg |
| Treasury bonds | Currency | yes — coupon pass-through or a promised rate | Duration: the pool can be worth less than face at any instant |
| Gold | Mass — grams or troy ounces, not currency | no | The currency price of the metal, which is the point |
Two consequences worth stating plainly:
The choice also decides the holder's experience at a till. A stablecoin- or bond-anchored instrument spends at face, like a note. A gold-anchored one needs a price to spend — which means either the till quotes it, or it is presented for redemption rather than payment. Both are workable; only one is a cash note.
Metals, bonds, currencies, equities and funds all fit the same pattern — a unit, a quantity, a reserve ledger — but they do not behave alike, and four difficulties appear as soon as the anchor is not cash:
| Anchor | Unit | Income | The difficulty |
|---|---|---|---|
| Currencies | Currency | Interest, if the issuer pays it | Trivial. This is the case the sandbox models. |
| Metals | Mass | None, and storage costs | Price risk in currency terms; assay and custody to prove the metal exists. |
| Bonds | Currency, plus a duration | Coupons | Value moves with rates; a face-value redemption needs a buffer. |
| Equities | Shares | Dividends | Corporate actions: splits and mergers change what a "share" is. |
| Funds / ETFs | Units of a fund | Distributions | Same as equities, plus creation/redemption mechanics and a manager. |
A two-for-one split doubles the share count. A token denominated in one share would then represent half of what it did yesterday, and its fixed denomination cannot be edited. The workable handling is the same as interest: keep the token fixed, record the action in the reserve ledger, and either issue the adjustment as new instruments or express it as a ratio applied at redemption. Mergers and takeovers are worse — the underlying ceases to exist, so the reserve must be redefined and holders told.
Coupons are scheduled; dividends are announced; fund distributions vary. An accrual plan that works for a bond (a known rate on known dates) does not work for an equity whose payment is declared quarterly with no guarantee. The schedule has to be data, not arithmetic in code.
Equities and funds settle a day or more after trade. A token that redeems instantly against a reserve that settles tomorrow is a funding gap, and it widens exactly when the market is stressed. Either redemption is delayed by the settlement cycle, or the issuer funds the gap and should price it.
This is the one that decides whether equity-anchored tokens are possible at all, in a given jurisdiction. A bearer instrument pays whoever presents it. Securities generally require a register of holders, transfer restrictions, and knowledge of who the holder is. Those two things pull in opposite directions, and no amount of engineering resolves it — it is a legal question that must be answered before the first token is issued.
Three positions a bank can take, in order of how much they change the design:
ESCROW_ONLY contract already expresses this.The bank contracts the value and CoinCard anchors it. A token can say "one hundred shares of this line", or "the currency value of one hundred shares on the redemption date", and the bank is the party that promise belongs to. CoinCard supplies the machinery that makes the promise signable, storable, transferable and checkable — not the promise itself.
That division has a precise technical consequence: the denomination has to be able to carry a quantity in a named unit, referencing a named thing, and all of it has to be inside the signed preimage. Today's canonical string carries an amount and a currency, which expresses cash and nothing else:
v1 COINCARD-ISSUE-V1|tokenId|amountCents|currency|contract
→ "1,000 cents of cUSD, peer-to-peer"
v2 COINCARD-ISSUE-V2|tokenId|quantity|unit|reference|terms
→ "100 shares of ISIN GB00B…, escrow-only, tracked"
→ "50000 cents of USD, p2p"
→ "10 grams of XAU, p2p"
unit is the thing being counted (share, gram, cent), reference is
which one (an ISIN, a metal code, a currency), and terms is the transfer condition
the bank attaches. Because all four are signed together, a token cannot be re-presented as a
different quantity of a different thing.
The sandbox signs v1 only. v2 is specified here because an equity- or metal-anchored token cannot be honest without it, and it is a contained change: a wider canonical string, two extra fields on the token record, and the signer and validator taught to carry them.
| Obligation | Whose it is |
|---|---|
| The promise the token represents | the bank |
| Holding the reserve, and proving it exists | the bank |
| Whether holders may receive it at all (securities law, transfer restrictions) | the bank |
| Corporate actions on the underlying | the bank |
| Making a token unforgeable, storable, transferable and verifiable | CoinCard |
| Recording who holds it, and refusing a spent copy | CoinCard |
| Keeping the physical anchor — the chip that holds the master copy | CoinCard |
That is a clean split, and it is the one worth putting in front of a bank: they are not adopting a currency, they are adopting an instrument format that happens to be auditable. The instrument is theirs; the guarantee that it cannot be copied or double-spent is ours.
For a metal-anchored instrument the contract has to say what happens at the end, and there are two honest answers. Which one is chosen is the bank's wording; what matters technically is that the choice is in the token, because a holder who can switch between them after the price moves is holding a free option against the issuer.
terms DELIVER_ITEM the holder receives the metal: bars, coins, a vault transfer
SPOT_VALUE the holder receives the currency value at a named price fixing
HOLDER_CHOICE either, at redemption, with a fee that prices the delivery option
| Delivery | Spot value | |
|---|---|---|
| What the holder gets | grams, in a form and place named in the contract | currency, at a stated price and time |
| What the issuer needs | custody, assay, logistics, an insurance chain | a price source, a fixing time, and a currency |
| The argument it invites | fineness, form, delivery cost, jurisdiction | which price, fixed when, and by whom |
| Spends at a till | no | no, unless quoted |
Two details that decide whether either version is trustworthy:
Neither settlement type belongs in the retail flow. A retail token is a contract for cash and spends at face; a metal token is presented for redemption, and the till does not care what the metal is worth that morning.
Retail tokens are contracts for cash. They are denominated in a currency, they spend at face, and they need nothing beyond what the sandbox already signs. The whole retail path — till, float, change, purse, sweep, offline card — is built on that one assumption, and it should stay that way.
| Retail | Investment | |
|---|---|---|
| What it is a claim on | Cash, in a named currency | Shares, bonds, metal, a fund |
| Denomination | Amount + currency (v1) | Quantity + unit + reference (v2) |
| Spends at a till | yes, at face | no — presented for redemption, or quoted |
| Transfer | P2P, escrow or redeem-only | Tracked only; the record always names the holder |
| Why | Cash has no register of holders | Securities generally require one |
The distinction is not cosmetic: it decides whether a token may move without asking anybody.
A retail note can be handed over because cash can. An instrument anchored in a security cannot,
because the law that governs it wants to know who holds it — which is precisely what the
ESCROW_ONLY contract already enforces.
Keeping them apart also keeps the retail path simple. Every feature a shop needs — float, change, purse, sweep — works on a fixed cash denomination and would have to be rewritten for something whose value moves. A bank that wants both runs two token classes from one account structure, not one token class with two behaviours.
| Decision | Why it cannot be deferred |
|---|---|
| Is the token a claim on a named pool, or on the bank generally? | A named pool allows a run on that pool; a general claim is a deposit and carries deposit-insurance and resolution questions. |
| Who takes the duration risk? | If instruments are redeemable at face on demand and the pool is long-dated, the bank is funding liquidity risk with someone else's certainty. |
| Is the yield promised or passed through? | A promised rate is a liability with capital behind it; a pass-through is a service. |
| What happens on a loss? | Bonds can fall. A token that always redeems at face needs a buffer, and the buffer is a capital decision. |
| Where does it sit in regulation? | Bond-backed instruments are plausibly securities; stablecoin-backed instruments are plausibly e-money. Both change who may hold them and how they are sold. |
The honest framing for a bank. A CoinCard instrument is a bearer claim, and what makes it worth its face is the reserve the issuer holds and the record that names the holder. Backing it with bonds adds yield and adds duration, credit and disclosure obligations. Backing it with a stablecoin adds a counterparty you do not control. Neither is free, and neither should be described to a holder as "the same as cash".
cUSD,
with P2P, ESCROW_ONLY or REDEEM_ONLY.reserve_id on the instrument, a
reserve ledger recorded by the issuer, and a periodic issuance that the vault signs.Sandbox deployment, USD, English. Test values, not money. This page describes a design space, not a product offering; the regulatory position of any reserve-backed instrument is a matter for the issuing institution and its counsel.