Why this needs Zama

The test to apply to any project that claims it needs a particular technology: delete that technology and see whether the product survives. If it still works, the technology was decoration.

#Delete the encryption and there is no product

Fully homomorphic encryption, usually shortened to FHE, means arithmetic performed directly on encrypted numbers, producing an encrypted answer, without ever decrypting the inputs. Zama's Protocol brings that to Ethereum: a Solidity contract can add, compare and choose between values it cannot read.

Take it out of Hearth and here is what is left.

What you get is PoolTogether. PoolTogether already exists, it works, and it has been running for years. There is no reason to rebuild it.

The product Hearth actually sells is the thing PoolTogether cannot offer: prize savings where your balance, your odds and your wins are yours alone, while the draw stays checkable by strangers. Those two properties are in tension on a transparent chain. Encrypted computation is the only thing that resolves them, and Zama's Protocol is the only place on Ethereum that does it today.

There is no partial version. Every one of the five rows above is a core promise. Remove the encryption from any one and the product fails at that row.

#The exact pieces we use

Not "built on Zama". Here is the list, with what each one does for us.

#Encrypted integers

euint64 for money and weights, euint128 for the pool's total accumulator, ebool for the outcome of a comparison. Every saver's principal, winnings, weight and credit is one of these, and so is each tier's carry. The arithmetic we perform on them is FHE.add, FHE.sub, FHE.mul by a public number, FHE.min, FHE.gt, FHE.le, FHE.and and FHE.select.

Comparison is doing more work here than the winner test alone. Five encrypted comparisons per draw place the pool's total weight against the powers of two around its last known bracket, and the only thing that leaves the encrypted world is the small count of how many of the five it beat. That is how the draw gets a public scale to run against without the total itself ever becoming a number.

FHE.select deserves a note, because it is what makes the whole design possible. It is an if-statement whose condition is encrypted: it returns one of two encrypted values and the chain cannot tell which. That is how a winner and a loser produce identical transactions. Nothing branches on a secret anywhere in Hearth.

FHE.fromExternal takes an encrypted value a user built in their browser, with its proof, and turns it into a value the contract can use. That is how a deposit amount arrives encrypted from end to end.

#ERC-7984, the confidential token standard

Each pool's asset is one of Zama's confidential tokens, an ERC-7984 wrapper around an ordinary ERC-20: cUSDC, cUSDT, cWETH, cBRON, cZAMA, ctGBP or cXAUt. Balances in them are encrypted values rather than public numbers.

Deposits arrive through confidentialTransferAndCall, which transfers an encrypted amount and calls the receiver's hook in the same transaction. The vault's hook is handed the amount the token really moved, which is how the vault credits reality rather than a request. Payouts go the other way through confidentialTransfer.

Using the standard token, rather than writing our own, matters. Several projects in this field hand-rolled an "ERC-7984-style" token. Every one of ours is a token Zama deployed, so a saver's confidential balance is usable outside Hearth and the token's own behaviour is not something we get to define in our favour. It also means Hearth can open a pool on a new confidential token the day Zama publishes it, which is how six of the seven were added, and that it can open none at all on a token whose mint the issuer keeps to itself.

#Encrypted randomness

FHE.randEuint64() generates a random number inside Zama's coprocessor, under the network's FHE key, from a seed that is public but useless without that key. The number comes out as ciphertext. Nobody, including us and including whoever sends the transaction, sees it at the moment it is created.

It has to be generated inside a transaction, because it mutates the on-chain generator state. That rules out the trick of previewing a draw off-chain with eth_call to see whether you would win, and it is why closing a draw is a real transaction that succeeds exactly once. Nobody can re-roll a seed they do not like.

#The access control list

Zama's on-chain ACL decides who may decrypt which ciphertext. It is enforcement, not policy: the relayer refuses a request for a handle the caller is not allowed on.

Hearth uses four calls on it. FHE.allowThis keeps a value usable by the contract in later transactions. FHE.allow grants a saver permanent read access to their own principal, winnings, per-draw weight and per-draw credit. FHE.allowTransient grants access for the length of one transaction, which is how the vault hands the pool a one-time allowance over an evaluation batch's total without ever giving it standing access. FHE.makePubliclyDecryptable opens a value to everybody, and we use it on exactly six kinds of value: the seed, the scale count that gives the bracket, the non-empty flag, the harvest, a tier's carry when that tier is due to reconcile, and the unfunded counter. The pool's exact total weight is deliberately not on that list.

That last call is one-way and permanent. It is the single most consequential thing a contract on this protocol can do, so every use of it in Hearth is listed in what stays private.

#EIP-712 user decryption

This is how a saver reads their own numbers. They sign a typed structured message, which is a signature standard that shows the signer exactly what they are approving, and Zama's relayer returns the plaintext of values that saver is allowed on.

It is an off-chain request. No transaction, no gas, no trace. That is why Hearth can have no claim function at all: learning that you won costs nothing and leaves nothing behind. The other half of that promise is that evaluation cannot be aimed at yourself either, so there is no transaction of any kind that only a winner would send.

Both the balance and the winnings are decryptable by their owner. Hearth also grants the per-draw weight and per-draw credit, so a saver can verify the draw's arithmetic against their own inputs rather than being asked to trust it.

#KMS-signed public decryption

The other direction. A contract marks a value publicly decryptable, anyone asks the relayer for the cleartext, and the relayer returns it with a signature from the key management service, the group that holds the network's decryption key. The contract then verifies that signature on chain with FHE.checkSignatures before acting on the number.

This is what turns "we say the seed was 12345" into a number the contract itself refuses to accept without proof. Hearth uses it once per draw for the seed, the scale count, the non-empty flag and the harvest together at award time, and again for a tier's carry whenever that tier is due to reconcile, which on Sepolia is every tier at every draw. Every proof is bound to its handles in a fixed order, so nothing can be shuffled or replayed into a different draw or a different tier.

#What a saver actually trusts

Naming this is the point of the page.

• The Zama Protocol to compute correctly on ciphertexts and to decrypt only what is marked decryptable. Every decryption the contracts act on carries a proof verified on chain. This is the same trust boundary Zama's own Confidential Vault documents.
• The confidential token wrappers, which are Zama's contracts rather than ours, and which are upgradeable by their owner. See the token layer section of what stays private.
• Hearth's own contracts, which are immutable once deployed, with no proxy and no upgrade path. The owner's remaining powers are narrow and listed in the threat model: a pause that stops deposits and draw closing but never withdrawals or evaluation, a yield-source setter, a rescue path for foreign tokens that cannot touch saver balances, and two-step ownership transfer with renouncing disabled.

Nothing on that list is a person we ask you to believe.