The Money That Can't Keep a Secret

The most transparent money ever built is about to learn some discretion.

Pay for groceries with a card and the merchant learns almost nothing about your finances. The terminal says approved or declined. Your balance, your salary, your other purchases stay where they belong, which is none of the merchant's business. Cash reveals even less.

Now pay from a stablecoin wallet on a public blockchain. The merchant has your address, and your address is an open book: your balance, every payment you've made, every payment you'll make next, readable by the merchant, their analytics vendor, and any stranger with a block explorer. Forever.

We don't talk about this enough. On the way to building the fastest, most programmable money in history, we also built the most transparent money in history. That transparency was a feature everywhere stablecoins grew up. It becomes a bug the moment they walk into a store.

It's the same irony this series keeps running into. Blockchains earned trust by making everything visible. Making them fit for commerce means learning what to hide.

Regular readers know the argument, so I'll keep it short. Card rails are built to interrupt. A payment mid-flow asks for a tap on a push notification, a one-time code, a 3-D Secure challenge, because the system reasonably assumes a human is standing at the checkout. For a human, that's a feature. For a delegated task it's a contradiction: you asked your assistant to reorder the groceries precisely so you wouldn't be pulled back in, and the payment step pulls you back in every time. Every approval prompt un-delegates the task. An agent needs money that settles in seconds, with finality, inside the limits its owner set, on any side of any border. That list describes stablecoins, which is why the agent economy is being built on them.

But bringing stablecoins to retail means bringing their defaults along, and one default doesn't fit.

Look at the customer's side first. Complete one purchase and the merchant can, in principle, look up the wallet that paid. How much is in it. What it spent last month. Whether it's a fifty-dollar wallet or a fifty-thousand-dollar one. Nobody has to abuse this for it to be wrong; a stranger being able to check your balance because you bought coffee beans is simply not a property money should have. And it does invite abuse. And a readable balance is a priceable one. Pricing has followed every other signal the internet made legible, and there is no reason to expect this one to be the exception.

The merchant's side of the glass is just as uncomfortable, and less discussed. A receiving address is a public revenue feed. Anyone, a competitor included, can watch the payments arrive and sum a month of sales. No business publishes its revenue in real time by choice.

On a public chain, accepting payment is publishing.

None of this is a scandal. It's an inherited default. Stablecoins grew up in DeFi, where radical transparency was the point: you audit the pool, you verify the collateral, you trust the math because you can see all of it. Commerce runs on the opposite instinct, refined over centuries. The transaction is between two parties, and its details are theirs.

Confidential transfers are exactly the problem Sui started addressing in June, when it launched the feature in public beta on Sui. The design is worth a close look, because it hides precisely the right things and nothing more.

Transfer amounts and account balances are encrypted on-chain, using twisted ElGamal encryption with zero-knowledge proofs (documented here), so the network can still verify that every transfer is valid, no overdrafts, no minting out of thin air, without anyone learning the numbers.

Who transacts stays visible. This is not anonymity money, and that's deliberate; compliance firms including TRM Labs and Merkle Science are exploring with Sui how risk scoring, monitoring and investigations work inside this model. For commerce, that's the correct trade exactly: the merchant sees your payment, not your balance. Observers see that a store has customers, not what its revenue is.

It's early, to be fair. This is a devnet beta, with testnet later this year. But the direction matters more than the date. The chains are learning discretion, and it's also why our sister company First Digital Labs is working on enabling confidential transfers for FDUSD on Sui, where FDUSD has been live since 2024. When agents carry money for millions of people, that money will need to keep their secrets the way cash always did.

Here's the part I keep coming back to, because it decides whether any of this reaches normal people. The general public is not ready for stablecoins and web3 wallets, and waiting for them to get ready is the wrong plan. They shouldn't have to get ready. Making that unnecessary is our job, not theirs.

In Hong Kong, where we're based, everyone carries an Octopus card. You top it up with cash or a card, then tap it on the bus, at the 7-Eleven, at the ferry pier. Nobody tapping one thinks about the settlement system behind it, and nobody should. It's the best payments lesson I know: the more invisible the rails, the more people use them.

That's what the consumer end of agent commerce has to feel like. A person sees value, theirs, in an app: available to spend, topped up when it runs low. An assistant they trust spends from it, within the allowance they set. Whether that value moves as FDUSD on Sui with encrypted amounts is our concern, the way the transit operator's back office is not the passenger's. Confidential transfers fit this picture, because privacy you have to configure is privacy most people won't have. It has to be the default, invisible, like everything else about good rails.

The earlier pieces in this series argued there's a person behind every agent ("Agents Don't Want Things"), and that the internet's walls are getting doors ("The Good Bot Problem") for the agents those people send. This one adds the third piece: the money the agents carry is learning to keep its mouth shut. Quietly, the oldest, most human properties of cash, instant, final, and none of your business, are being rebuilt for machines. That's the whole project, in one line.

Third in a series on the agent economy, after "Agents Don't Want Things" and "The Good Bot Problem."

I'm the CTO at First Digital, building payment infrastructure for the autonomous agent economy. Tell me where I'm wrong on LinkedIn or X.

To learn more about Finance District, join our Discord community, follow us on X, or explore our documentation.