> Who cleans up the code? SETCODEFROM and the bytecode nobody points at
Ethereum · EIP-8298 · EIP-8297 · state · 2026
Every client already keeps bytecode no account points at, and it never mattered. SETCODEFROM lets any transaction add to the pile, and the partitioned binary tree moves code into the state root. Where dangling code comes from, client by client, and three ways to ship SETCODEFROM.
> What your account can become after SETCODEFROM
Ethereum · EIP-8298 · EIP-7702 · account abstraction · 2026
A diagram-first tour of every working mode SETCODEFROM opens up: keep your ECDSA key while going code, cheaply deploy a copy of an existing contract, retire the key for good, or come back to ECDSA later, with the exact EIP behind each door.
> Pay like a card, settle like Ethereum
Ethereum · FCR · EEZ · rollups · payments · 2026
Card-terminal UX for payments from any rollup, with no bridge, no middleman and no one new to trust. The Fast Confirmation Rule and the Ethereum Economic Zone, stacked, let a shop that only watches Ethereum ship in about two slots today, and in a card tap's time as slots shrink.
> Why two identical Geth benchmarks disagree by 7×
Ethereum · geth · pathdb · benchmarking · 2026
Identical EEST bloatnet runs report wildly different MGas/s on three geth databases. The cause turned out to be a 380 MiB pathdb journal shipped inside one snapshot — a provenance artifact, not a property of either database.
> How can two Besu databases holding the same state differ 9×?
Ethereum · Besu · Bonsai · RocksDB · benchmarking · 2026
The same experiment on Besu. Three findings, all about how the stores were built: the pre-run's rows in the snapshot's youngest files, a generated store with no bloom filters, and a code pool unlike mainnet's. Fixed, absence lands at parity and shared code within 1.5%; distinct-contract code overshoots by an amount that depends on the disk.
> How can two Nethermind databases holding the same state differ 17×, and then agree?
Ethereum · Nethermind · flat DB · RocksDB · benchmarking · 2026
A mainnet snapshot and a generated store, the same benchmark suite, up to 17× apart. Five things about how the stores were built and measured made the gap; fixed one at a time, the tests that run over a second land at parity on two independent pairs.
> How can two reth databases differ 64× on the same benchmarks, and then agree?
Ethereum · reth · MDBX · benchmarking · 2026
A mainnet snapshot and a generated store, the same benchmark suite, up to 64× apart. A buffered loop, an older storage layout, and four things in how the store was generated. With all of them fixed, 153 of the 154 tests that run for half a second or more land within ±10%.