A quiet Saturday overall, but post-quantum transaction costs got their first real measurements. The proof batching draft was assigned EIP-8500, and its author Marchhill used benchmarks to argue for cutting leanSPHINCS verification gas from 15,400 to 3,000 and leanSTARK from 572,000 to 30,000 (all provisional). EIP-8151's public mempool rules got a reviewer's approval, and ERC-8499 discussion moved on to mnemonic fingerprint design. The EIP-8037/8038 repricing audit findings land at ACDT #103 on the evening of 11-16.
Image: Default image (no real image available for this day)
Post-quantum transaction costs: EIP-8500 benchmarks and proposed cuts
Between 14:22 and 22:35 on 2026-11-14, author Marchhill pushed 7 revisions, including "simplify batching to ordinary generic dependencies", "separate mempool verification budgets from gas calibration" and "bound block dependencies independently of the gas limit". At 06:17 on 2026-11-15, editor abcoathup suggested eip: 8500 in a review, assigning the official number.
Was updated again at 03:33 on 2026-11-15 and is still unmerged. Which post-quantum signature Frame transactions adopt determines the real load on EIP-8500.
At 02:59 on 2026-11-15, colinlyguo suggested strengthening the rationale: for the "any other digest" case the risk isn't just signature reuse. Given a known public key and an attacker-chosen digest, an attacker can build a (digest, v, r, s) tuple that recovers the target address without the private key. At 03:08 the review concluded "lgtm, only a rationale supplement". The PR isn't merged yet.
At 16:46 on 2026-11-14, Helkomine argued against adding features to EIP-7702: its goal is to move EOAs gradually toward smart accounts, and ECDSA will be phased out in a post-quantum world, so there's little reason to keep hardening it.
At 22:17 on 2026-11-14, thomasharris noted that eth_getProof can prove an EIP-7702 account's codeHash (that is, 0xef0100||address) at a given block, but redelegation emits no event. At 01:01 on 2026-11-15, Zealynx added that an account can switch away and back within one block, so scanning codeHash per block can miss it. Comparing the nonce at both ends of the period is more reliable, since every applied authorization bumps the nonce by one.
At 16:16 on 2026-11-14, SergeevDmitry confirmed the author had added "keep deriving old key types" to the restore section. At 01:04 on 2026-11-15, Zealynx suggested fixing the fingerprint format and computing it from a public key (for example account 0's SLH-DSA-SHA2-128s key), the way BIP-32 takes its fingerprint from the master public key. If it's computed straight from the seed, anyone holding the paper with the fingerprint can test passphrase guesses offline at PBKDF2 speed. For stateful keys he backs picking leaves at random, because a device reset or a rolled-back flash makes a counter repeat.
Updated at 16:30 on 2026-11-14. At 21:00 on 11-16, MariusVanDerWijden gives the EIP-8037/8038 status, ChainSecurity and Dedaub present independent audit findings, and the EF covers its internal investigations and app team outreach, including compatibility PoCs for the Juicebox protocol. The call also picks one of two consensus-specs PRs for payload binding, to unblock the consensus spec release.
At 06:00 on 2026-11-15 it listed ERC-8489 through 8494 and others, mostly about AI agents (information flow, settlement deductions for task escrows and so on), plus the ERC-8493 onchain URI scheme. ERC-8489 and ERC-8490 come from the same author, garyyang-finchip. There are a lot of agent ERCs, so check how mature the discussion is before building on one.
Vitalik: make monitoring systems as un-agentic as possible a reply at 11:28 on 2026-11-14, followed by a pointer to an open and verifiable hardware project he's funding at 11:43.