Ethereum Draft Proposes Committee of About 512 Validators and Separate Voting to Speed Finality
Key Facts
Ben Edgington opened a pull request on 21 October 2026 to add a draft change to Ethereum's consensus mechanism, envisaging a committee of about 512 validators voting on the chain head in each slot. The design is part of the "decoupled consensus" approach, which separates validators' votes to select the chain head from their votes on finality, according to his explanation published by the Ethereum Foundation's consensus team. Votes for each track propagate across the network and are processed independently of those for the other track. According to the draft in the Ethereum Improvement Proposals repository, a validator's probability of selection for the chain-head committee is proportional to its effective balance. The pull request remained open when checked, in a step aimed at changing the voting process that precedes transactions receiving a finality guarantee.
Edgington estimates the expected average time for a transaction to reach finality under the current design at about 960 seconds, while Ethereum's official website puts the time for a block to reach finality at about 15 minutes. The website defines finality as a guarantee that prevents a block from being changed or removed without burning at least 33% of all staked ether. In the current timing structure described by Edgington, an epoch consists of 32 slots. Each slot lasts 12 seconds, under the existing mechanism for operating the network and voting. During normal operation, the number of recent blocks that have yet to reach finality ranges from 63 to 95, according to his account of the current consensus mechanism.
The cost of waiting stands out in Edgington's research into ecosystem needs, which drew on interviews with 19 individuals and organizations, as well as written feedback from other participants. According to the research, transfer executors and intermediaries fund liquidity upfront until settlement, and Edgington wrote that "every minute of finality delay ties up capital." Participants frequently described finality of between 10 and 20 seconds as a suitable range for bridges, interoperability systems and systems that execute user intents. Networks built on Ethereum also reported that slow finality leaves their transaction pricing reliant on stale information about data costs on the underlying network. In an existing cross-chain transfer application, Circle's documentation explains that fast messages become eligible for attestation at the confirmation level. Standard messages are attested at the finality level, according to the company's technical documentation.
On security, the draft seeks to limit the influence of old votes on chain selection. It aims to prevent attackers from accumulating withheld votes and then releasing them strategically to influence the selection process, according to the proposal. In validator operating economics, the design does not seek to change the overall budget for rewards they receive for participating. It does, however, allow the distribution of rewards across voting duties to be adjusted, while keeping the total unchanged at the same staking level. This coincides with another finding from the ecosystem needs research: participants showed little willingness to improve performance by imposing costs that would exclude home validators from participating.
The design also paves the way for subsequent changes to signing and vote aggregation, but does not select a quantum-resistant signature scheme, according to the draft. Implementation requires a consensus-layer hard fork, alongside updates to consensus and validator software to support the new voting duties. The required updates also cover block and state formats and fork-choice rules, tying implementation to changes in the components that operate the consensus layer. The Ethereum Foundation identifies decoupled consensus as a leading candidate to be the centerpiece of the "I-star" upgrade, with work under way on its targeted specifications and prototypes. In the pull request to add the draft, Edgington said "many aspects remain unspecified" and that the main purpose at this stage was to obtain a proposal number. In his research vision, he described including decoupled consensus in that upgrade as an ambitious goal. The upgrade's possible timing during 2027 remained part of that vision, rather than a confirmed implementation date.