Ethereum Researchers Pursue Redesign to Cut Finality to Seconds
The proposed architecture would separate block production and fork choice from finality, with smaller committees handling some consensus functions.
Ethereum protocol researchers are pursuing a consensus redesign that could reduce finality from 13 to 19 minutes to seconds, potentially accelerating cross-chain swaps and Layer 2 deposits while lowering capital costs for bridges and solvers.
The proposed architecture would separate Ethereum’s available chain, which continues producing and selecting blocks, from a finality mechanism that confirms checkpoints. Smaller committees of about 256 to 512 validators would handle block production and fork choice, while separate votes would finalize the chain.
Ethereum’s current design uses attestations from about 30,000 validators per slot to support fork choice. The proposed separation is intended to reduce the work required for that process while preserving a security mechanism that makes finalized blocks difficult to reverse.
“First, we will decouple finality from the rest of the consensus process,” Ben Edgington, Ethereum Foundation Consensus, wrote. He also called decoupled finality “the big unlock.”
Ethereum’s normal finality time is 13 to 19 minutes, with current finality taking about 15 minutes. Reversing a finalized block would require an attacker to burn at least 33% of staked Ether (ETH).
The faster-finality program is designed to support cross-chain swaps and Layer 2 deposits, reduce application risk-management requirements and lower capital costs for bridges and solvers. It is also intended to shorten the period during which transactions remain exposed to attacks before finalization.
Specifications and prototypes are underway. Decoupled consensus is the leading candidate for the I* fork, but no finality upgrade has been assigned to a specific fork.
Related work includes single-slot finality, three-slot finality and Minimmit. The Etheorem project’s Lean 4 consensus specifications passed official test vectors for the Fulu, Gloas and Heze forks. That result does not establish that the proposed decoupled-finality protocol itself has been formally verified.
