πŸ“‹ Thread 2 β€” Block-Level Access Lists. Yesterday: gas got two dimensions. Today: every block commits a list of *what state it touched*. EIP-7928. The Glamsterdam EL headliner. 🧡
β›½ Thread 1 β€” the 21,000-gas myth. A plain ETH transfer was always 21,000 gas. On Amsterdam, that number still exists β€” and it will still brick wallets. Here's why. 🧡

Aug 26, 2026 Β· 11:13 AM UTC

1
2
1,008
Status quo (today): Execute a block β†’ txs run β†’ state changes. Which accounts were read? Which slots changed? You discover that *during* execution β€” or by replaying traces. Nothing in the header promises it upfront.
1
39
Amsterdam adds a Block-Level Access List (BAL). After the block runs, the VM records which accounts were accessed β€” balances, nonces, code, storage. Beside the block body, not inside. Header: blockAccessListHash. eips.ethereum.org/EIPS/eip-7…
1
22
Why it matters: Validators and builders can see state touch patterns *before* parallel execution work gets serious. Glamsterdam pairs this with ePBS on the consensus layer. EthereumJS covers the execution side β€” run it locally.
1
12
Builder flow in @ethereumjs/vm: runBlock(vm, { block, generate: true, }) No extra flag beyond Amsterdam. The VM accumulates accesses automatically while executing.
1
12
Then read: result.blockLevelAccessList One plain ETH transfer β†’ 4 accounts in the BAL on our example. Header hash must match bal.hash(). Mismatch = invalid block.
1
10
Validator flow: runBlock(vm, { block, blockAccessList: balJson, }) Pass the known list (JSON, RLP, or BlockLevelAccessList). VM checks structure + hash, then compares against execution.
1
12
The gotcha: BAL is not per-tx β€” it is per *block*. Use Hardfork.Amsterdam (full bundle), not EIP-7928 in isolation. Peers exchange it separately (eth/71). Your indexer cannot ignore the header hash.
2
23
Full walkthrough + validator example: runBlockBALValidate.ts Docs: github.com/ethereumjs/ethere… Experimental β€” not production. Next: ETH transfer logs (EIP-7708). πŸš€ [ END OF THREAD ]
27