Glamsterdam on Sepolia
Every ETH transfer now leaves a log. What that changes for stealth payments, measured on Sepolia since the fork on 6 October 2026.
Glamsterdam went live on Sepolia on 6 October 2026 at 13:53:36 UTC, from block 11,856,337. Two of its changes matter for stealth addresses: every ETH transfer now emits a log (EIP-7708), and every block lists the accounts it touched (EIP-7928). Mainnet has no date yet (EIP-7773). We read Sepolia to see what changes before it reaches mainnet.
What we read
Every ERC-5564 announcement on Sepolia from the fork block to block 11,957,988 (read on 22 October 2026), and every EIP-7708 log to or from the stealth addresses they announced. Numbers only, no addresses: the same rule as our leak study.
| Sepolia since Glamsterdam | |
|---|---|
| Stealth payments announced | 19 |
| Stealth addresses that received ETH | 6 |
| ETH received | 0.4112 ETH |
| Addresses that sent ETH out | 5, in 8 transfers |
| Of those, sent ETH back to an address that paid them | 5 |
The other 13 addresses have no ETH log at all in this window: no ETH reached them.
Cheaper to read
Before Glamsterdam, a wallet learns what happened to the ETH at its stealth addresses one address at a time: a balance read and an explorer query per address. ETH that a contract moved out (an internal transfer) shows up only in traces, which public nodes rarely serve.
After it, two eth_getLogs filters cover any number of addresses: Transfer logs emitted by the system address, with your addresses as recipients (ETH in) or as senders (ETH out). For the 19 addresses above that is 2 requests per block range instead of at least 38, and internal transfers are included.
const SYSTEM = '0xfffffffffffffffffffffffffffffffffffffffe';
const TRANSFER = '0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef';
const pad = (a: string) => '0x' + a.slice(2).toLowerCase().padStart(64, '0');
// ETH in: topic 2 is one of your stealth addresses. ETH out: topic 1 is.
const ethIn = { address: SYSTEM, topics: [TRANSFER, null, stealthAddresses.map(pad)] };
const ethOut = { address: SYSTEM, topics: [TRANSFER, stealthAddresses.map(pad)] };Always check the emitter. Any contract can emit a Transfer event with the same topics; only the system address cannot be faked. And the logs show where ETH went, not the gas paid, so read the balance itself when you need it.
Cheaper for everyone
The same two requests work for anyone watching. With a plain node, every announced stealth address can be followed to where its ETH went, no explorer and no traces needed. On Sepolia, all 5 addresses that sent ETH out sent some of it back to an address that had paid them:
- one sent everything back to the address that paid it and posted its announcement, the round trip our H2 check blocks;
- four were paid by the same contract and each sent part of it back there, which ties all four together.
Withdrawal habits matter more after Glamsterdam, not less.
Paying a new address costs more
Every stealth payment goes to an account that does not exist yet, and Glamsterdam charges more for creating state (EIP-8037). In this sample, both payments that sent ETH straight to a new stealth address used 204,600 gas. A plain transfer out of a stealth address to an existing address used 21,000. Before the fork both would have used 21,000. Expect an ETH payment to a new stealth address to cost about ten times the gas of a plain transfer once Glamsterdam reaches a network.
In the client
Client 0.2.6 reads these logs. The collector check (H3) already read Transfer logs from the node, to catch token transfers the explorer has not indexed yet. On Sepolia every ETH withdrawal now logs a Transfer too, so H3 would have called the history incomplete after any ETH withdrawal. These logs now count as what they are: ETH withdrawals read straight from the node, including ETH a contract call moved out, which an explorer's transaction list does not show. Nothing changes on mainnet until Glamsterdam reaches it.
Not measured yet
Block-level access lists (EIP-7928) name every account a block touched. A scanner could use them to skip blocks that touch none of your addresses, without reading logs at all. We have not measured that on Sepolia yet. It is our next step.
Repeat it
On any Sepolia node: read the ERC-5564 announcements from block 11,856,337, then run the two filters above over the same blocks with the announced stealth addresses. The fork time and block are in EIP-7773.
Stay unlinked.