At Jackson Hole on 28 August 2026, the general manager of the Bank for International Settlements, Pablo Hernández de Cos, said stablecoins are not a credible means of payment at scale. His argument was not the usual hand-waving about volatility or crime. It was a checklist: stablecoins fail on par redeemability, elasticity, interoperability and financial integrity, and tokenised deposits make the better case for day-to-day payments.
Crypto’s reflex is to dismiss the man from the central bankers’ central bank. Do not. The diagnosis is largely right. It is the prescription that should worry you, and there is a version of this problem sitting inside almost every crypto gaming platform, because the proof everyone is proud of does not reach it.
TL;DR
- BIS chief Pablo Hernández de Cos told Jackson Hole on 28 August 2026 that stablecoins fail par redeemability, elasticity, interoperability and financial integrity, and pushed tokenised deposits instead.
- Par redeemability means one unit converts to one dollar on demand, always. That is a property of an issuer’s balance sheet, not of a blockchain.
- A Chainlink VRF proof covers who won, and a contract covers whether the transfer happened. Neither says anything about what the transferred unit is worth when you try to leave.
- This is the thirty-eighth unasked half of fairness: the platform picked your prize’s unit of account, and you never got a vote on the issuer standing behind it.
- The BIS cure, tokenised bank deposits, fixes par by adding a permissioned issuer with a state backstop. Real fix, bad trade for a permissionless game.
Par redeemability is a promise, not a number
Par redeemability sounds like plumbing. It is the whole thing. It means one unit of the token converts to one unit of the underlying currency, on demand, in full, without a queue, a haircut or a phone call. Bank deposits get this from deposit insurance, central bank liquidity and a supervisor who can force the issue. A stablecoin gets it from the issuer’s reserves, its redemption terms, its banking partners and its willingness to honour a claim from you specifically, at the size you hold, on the day you ask.
Notice what none of that is. It is not on-chain state. You can read a token balance from a contract with total certainty and learn nothing about whether that balance redeems at par. The number on the chain is exact; the promise behind it sits in a jurisdiction, in terms most holders have never opened, and in a reserve you get to inspect four times a year, which we have covered before: an attestation is a point-in-time, scope-limited opinion, not a proof.
Where this lands in a game
Take a raffle with a 10,000 unit stablecoin pot. What is genuinely verifiable: the ticket count is contract state you can read before you buy; the winner comes from a Chainlink VRF word whose proof the contract verifies before it will accept the number; the index is arithmetic anyone can rerun; and the payout moves in the same transaction as the selection, so no human approves it, delays it or reverses it. That chain of custody is airtight, and it is the reason on-chain gaming exists.
Now look at what is not covered. The word “10,000” has a unit attached, and that unit is an unsecured claim on a company. The VRF proof does not extend to it. The escrow does not extend to it. The contract can guarantee you receive exactly 10,000 of the thing; it cannot guarantee the thing is worth 10,000 of anything else. Your prize is exact in units and only approximately exact in value, and the approximation is set by a party you never chose.
That is the unasked half. Not fraud, not a hack, not an operator with a hand on the scale. Just a decision the platform made on your behalf before you arrived: which issuer’s promise your winnings would be denominated in. Nobody asks players this. Nobody discloses it as a risk. It shows up in the interface as a dollar sign.
Elasticity bites at the worst possible moment
De Cos’s second failure is the sharper one for anyone holding a balance. Elasticity is the ability of the money supply to expand on demand to settle obligations. Bank money has it. A stablecoin does not: the issuer must sell reserve assets to meet redemptions, and reserve assets have market prices and settlement windows.
In calm conditions this is invisible, because almost nobody tests it. It becomes visible in exactly the conditions where you want your winnings out: a market-wide scramble, a banking partner wobble, a redemption queue where institutional holders with direct agreements are served first. Retail does not redeem with the issuer at all. Retail sells on a secondary market at whatever the book says that minute. Par is a promise made to wholesale counterparties and inherited by everyone else on trust.
Every other failure mode in this industry has a mechanical answer: verifiable randomness for a rigged RNG, atomic settlement for a payout queue, immutable contracts for an admin key. Issuer risk has none available to the platform, because the failure is not in the code.
Why the BIS cure is worse than the disease, for us
Here is where the speech goes wrong for anyone building permissionless. De Cos’s answer to a fragile promise is a stronger promise: tokenised deposits, issued by regulated banks, par-redeemable because deposit insurance and a central bank stand behind them. He is right that this fixes par. He does not dwell on the price. Tokenised deposits are permissioned by construction: the issuer knows who you are, can decline to serve you and can freeze the balance. Swap a stablecoin for one in that raffle and you have traded an unverifiable promise for a verified gatekeeper. The prize is now genuinely worth what it says, provided the bank agrees you should have it. For a game whose entire premise is that no party can intervene between the draw and the payout, that reintroduces the exact veto we spent a decade removing.
What a platform can honestly do
Denominate in an asset that is not a claim. Native ETH is volatile, and volatility is a different problem from counterparty risk. An ETH pot can be worth less tomorrow, but no issuer decides that, and there is no redemption queue and no terms of service. For a game settled in a single transaction, price risk is the lesser evil.
Name the issuer in contract state. If a pot is denominated in a stablecoin, the token’s contract address is already in the pot’s on-chain state. Surface it instead of rendering the prize as an abstract dollar sign. Players who care can check which issuer, which chain, and whether the contract has a freeze function. Players who do not care were at least not misled.
Stop calling a token balance a dollar amount. “10,000 USDC” and “$10,000” are different claims. One is arithmetic over contract state; the other is a forecast about a company’s solvency and redemption policy. Interfaces conflate them because it reads better. It also quietly hides the only part of the transaction that is not verifiable.
None of this makes issuer risk go away. It converts an unasked question into an asked one, which is the entire project here. Provable fairness is not a badge certifying that everything is fine; it is a boundary with a clearly marked edge. The draw is proved. The payout is atomic. The unit is somebody’s word, and you deserve to know whose.
Satoshie runs raffles and coinflips on Base with Chainlink VRF, where the odds are contract state you can read before you buy and the payout settles in the same transaction as the draw. We will keep telling you which half of that sentence the proof covers.
📷 Photo by David Trinks on Unsplash
