A team of five Ethereum consensus researchers, including Francesco D’Amato and Ben Edgington, put up a draft Core EIP on Wednesday (Oct. 21st) called Decoupled Consensus, which would retire Gasper, the beacon chain’s current consensus design, and split validator voting into two separate jobs.
The payoff, if it ships, would be faster finality, tougher reorg defenses, and a cleaner onramp to post-quantum cryptography.
What’s the Scoop?
Two jobs, not one: Today, each validator attestation does double duty, voting on the chain’s head and counting toward finality. Under the draft, a committee of ~512 validators sampled each slot would handle head votes, while every active validator votes on finality in separate rounds that no longer wait for epoch boundaries.
Enjoying this article?
Subscribe to Bankless or sign in
The finality target: The EIP’s initial goal is at least a 4x cut in time to finality. Blocks currently need two epochs to finalize (~13 minutes at best), so on today’s numbers that’s roughly three minutes. Longer term, Edgington wrote on the EF’s consensus blog in July that finality “in a few tens of seconds” looks like a useful target.
Reorg defense: LMD-GHOST, Gasper’s fork-choice half, never expires old votes, so attackers can sit on votes and drop them strategically to sway the chain. The new design only counts recent votes, adds expiring locks to protect confirmed blocks, and replaces proposer boost with a vote-sharing mechanism called view-merge.
The quantum angle: The draft doesn’t choose a post-quantum signature scheme. Instead, it separates the voting paths so each can change how votes get signed and aggregated later. That matters because PQ signatures are far bulkier than today’s BLS signatures, and limiting per-slot head votes to ~512 signers, while giving finality votes more time to propagate and aggregate, leaves room for that extra weight.
Early days: Many pieces are still TBD, including committee selection, rewards, and the full safety analysis, and the authors say they’re mainly after an EIP number for now (a reviewer has proposed slotting it in as EIP-8453). The draft also flags that a small committee makes for an easier denial-of-service target. Shipping it requires a consensus-layer hard fork, and EF researcher Barnabé Monnot has pegged it for I*, the upgrade after Hegota, in early 2028.