Account abstraction is having another moment. Several outlets this week reported that Ethereum core developers had locked it into Hegota, the network's next upgrade after Glamsterdam. The framing is off in one important way. A working version of account abstraction has been running on Ethereum since March 2023, and the native version the headlines describe is not scheduled at all yet.
Account abstraction shipped in 2023, as an overlay
ERC-4337 went live in March 2023. It never touched Ethereum's consensus rules. Instead it sat on top of them, routing user actions through bundlers and a shared EntryPoint contract, which let wallets offer features that plain Ethereum accounts could not. By early 2026 that standard supported tens of millions of accounts and hundreds of millions of transactions, and it powers wallets such as Safe, Argent and ZeroDev. Pectra added EIP-7702 in 2025, a lighter bridge that let ordinary accounts borrow smart-account behavior for a single transaction.
So the capability is not theoretical. Teams shipped it, audited it, and run it daily, though adoption is uneven. Earlier this year more than 200,000 smart wallets went idle even as new ones kept launching. What none of it is, is native. Every part of the design lives in application code rather than in the protocol.
EIP-8141 moves the logic into the base layer
EIP-8141, called Frame Transactions, is the attempt to make account abstraction part of the protocol instead of a layer bolted on top. Vitalik Buterin has championed it personally. It would let an account pay gas in an ERC-20 token instead of ETH, authenticate with passkeys or multisig, batch several actions into one transaction, and recover access without a single seed phrase. Anyone who has tried to recover a lost wallet understands why that last one carries weight.
| Aspect | ERC-4337 (2023) | EIP-8141 |
|---|---|---|
| Where it lives | Overlay contracts | Protocol layer |
| Transaction unit | UserOperations, separate mempool | Frame Transactions, public mempool |
| Validation | EntryPoint contract | Native consensus rules |
| Gas payment | Paymaster contracts | Native payment frames |
Moving into the protocol is exactly what makes it slow. An overlay can be upgraded by the teams that maintain it. A consensus change has to be implemented and tested by every execution client at once, because native account abstraction cannot be switched on gradually. Developers have flagged unresolved questions in the transaction-pool design and disagreement over verification details. One research note this year called the feature too "heavy" to headline the fork for that reason.
FOCIL is scheduled. Account abstraction is not.
The scheduling is where the two stories separate. At the All Core Developers call on August 27, the only proposal with Scheduled for Inclusion status was FOCIL, a censorship-resistance change numbered EIP-7805 that lets a validator committee force builders to include eligible transactions. Frame Transactions sat a tier below, at Considered for Inclusion, still weighed against a rival account-abstraction design, EIP-8130, that is heading to Base first in September. One report dated the following day said EIP-8141 had cleared the bar to Scheduled for Inclusion after that call, though that account is so far single-sourced.
Even the optimistic reading does not produce a date. Scheduled for Inclusion means a proposal is in scope for a fork, not that the fork has a slot. Hegota is the upgrade after Glamsterdam, and Glamsterdam itself slipped from the first half of 2026 to the fourth quarter. Client teams do not submit ranked preferences until September 10, and Ethlabs' own priority list places account abstraction third, behind FOCIL and faster block times. Hegota is penciled in for 2027.
For anyone holding a smart wallet today, little changes in the near term. The overlay version keeps working, wallets keep shipping features on top of it, and the native rewrite stays a research target with a moving date. The open question is not whether account abstraction reaches Ethereum, since it did in 2023, but whether the protocol-level version arrives as one clean standard or splits between mainnet and Base. The client rankings due in September are where that starts to resolve.
