Future of Account Abstraction
Ethereum Korea One: Genesis · Day 2 · 2026-09-29 · 30:09 · In English
With ERC-4337 live for over three years and EIP-7702 for over a year, why is native account abstraction back on the table? At Ethereum Korea One: Genesis, the Geth lead and ZeroDev's tech lead compare the L1 and L2 paths.
Speakers
- Felix Lange, Geth Lead, Ethereum Foundation
- Taek Lee, Tech Lead, ZeroDev
- Seungmin Jeon, Core Engineer, Sunnyside Labs (moderator)
About this session
Moderated by Seungmin Jeon of Sunnyside Labs, this panel compares frame transactions (EIP-8141), the general mechanism being prepared for Ethereum L1, with EIP-8130, proposed by Base and supported by several L2s, after talks to merge the two standards collapsed two weeks before the event. Felix Lange of the Ethereum Foundation and Taek Lee of ZeroDev (now part of Offchain Labs) discuss why L1 and L2 need native account abstraction for different reasons, what changes when both standards coexist, and what comes next for wallets.
Key takeaways
- Felix Lange said ERC-4337 was a testbed for smart accounts, but at the protocol level transactions can still only originate from a key. EIP-7701, which would bring the 4337 environment to mainnet, was judged too prescriptive of one transaction flow and "not really in the spirit of Ethereum," so the team moved to the more general frame transactions. Vitalik pushed the initial idea, and Lange's colleague Matt Garnett now leads the work.
- Taek Lee said L2s need native AA for two reasons: post-quantum migration, which "we cannot actually delay anymore," and user experience, which L2s prioritize more than L1. Despite embedded accounts and session keys, most users have still not migrated off their EOAs, and 7702 is used for little more than batching.
- Lange noted that EIP-7702 lets an account deploy code, but the original key can still send ETH, so "it's not like a real smart account." The protocol should provide general primitives and leave implementations to companies like ZeroDev, while compressing requests such as native batching and transaction assertions into a minimal change that can ship in a fork.
- Lange described frame transactions as "a superset of 8130," able to do "pretty much everything 8130 can do and more." Much of the work has gone into builder-sponsored gas flows and aggregating post-quantum signatures, since every transaction cannot carry about 4 KB of signature data.
- Lee said L2s back EIP-8130 because time to market is critical and they did not want more fragmentation. He was once worried about wallets needing per-chain routing, but now expects little overlap: 8141 for full asset management and privacy, 8130 for embedded account UX. Lange added that an 8130-style key registry could exist on L1, but since L1 has no operator, it should not require a hard fork for every new signature type.
- Native AA can ship on mainnet at the H-star fork at the earliest, about a year out, so Lange saw Base shipping its own standard as understandable. He is most excited about leanVM and hash-based signatures ("we're going to be good forever"), while Lee is betting hard on the next generation of wallets built on frame transactions.