Native account abstraction slips again at ACD#243, and Poseidon is out

At All Core Developers Execution Call #243, developers finalized Glamsterdam's gas repricing and moved its public testnet, Plataberget, forward, but the decision on whether native account abstraction becomes part of Hegota was pushed back again ^[1][2]. Separately, the Ethereum Foundation dropped its Poseidon hash in favour of standard hashes ^[3]. Two roadmaps that teams have been planning against just moved: anyone waiting on protocol-level accounts, and anyone building proving infrastructure around Poseidon.

The awkward part is that client support went the right way. Nethermind now backs EIP-8141 Frame Transactions, the native AA enabler, and could champion it as a headliner; Geth contributors describe it as increasingly important for post-quantum readiness; Besu raised no opposition ^[4]. The brake came from Layer 2s. Base and Arbitrum objected to adopting Frames before the wider EVM ecosystem converges on a single account model, warning it could fragment the stack if L1 and L2s head in different directions ^[5].

That is not an abstract worry, because there is a rival design already shipping. Base has implemented parts of EIP-8130, which adds programmable-account features such as key rotation and cross-chain account sync without new EVM opcodes ^[6]. Frame Transactions supporters argue their concerns can be handled through further specification work ^[7]. Ethlabs still ranks Frame Transactions as its preferred path to native AA in its Hegota priorities paper, and the proposal has advanced to "Considered for Inclusion" ^[8]. Breakout calls resume on August 25, with a possible decision on August 27 ^[9]. Since Glamsterdam's scope is being closed out and native AA is only a Hegota candidate, the earliest protocol-level account model lands is Hegota, and that is contingent on the August calls going one way rather than the other ^[10].

So ERC-4337 remains the load-bearing option, which is roughly what its own retrospective concludes. ZeroDev's CTO, who led development of the Kernel smart account, published a three-part review after ERC-4337 reached Final status, arguing the standard succeeded but not as the consumer-wallet revolution many expected ^[11]. Across roughly 1.2 billion UserOperations, sponsored transactions made up the vast majority of volume, activity concentrated on a few deployments such as World Chain and Base, and there was little sign of users moving their main EOA into a standalone smart account ^[12]. The capability that mattered across Kernel v1 to v4 was replaceable validation, not batching or sponsorship; fine-grained permissions shipped as a scalpel and were mostly used as an on-switch ^[13].

His third piece argues the real unlock is in-place migration: change the authorization model, not the address, which is what EIP-7702 starts and native AA would finish ^[14]. The hard part he flags is deletion. After delegation, the original ECDSA key can still authorize changes, and many applications still trust a raw signature through ecrecover, so a genuine transition has to fix both protocol and application layers ^[15].

The adoption numbers say the same thing from the other end. In an AA Mafia Telegram discussion that pulled in people from Base, the Ethereum Foundation, ZeroDev, Etherspot and Ambire, one participant shared 7702 activation counts from roughly a week earlier: Trust Wallet over 640,000, MetaMask around 280,000, and Pimlico's Simple7702 account near 200,000, with more behind ^[16][17]. That is about 1.12 million activations across the top three ^[18], and most of those wallets do not enable 7702 by default, so a large base of plain EOAs remains ^[19].

Watch the August 25 breakout and the August 27 call for whether Frames gets a headliner slot or another deferral ^[9], and whether Base's EIP-8130 work hardens into a competing standard rather than a negotiating position ^[6]. If you have proving infrastructure keyed to Poseidon, the hash decision is the more immediate re-plan ^[3].