EIP-8390 Removal of the sync committee (Remove the Sync Committee)
No call decision · Consensus & staking · EIP status Draft · created 2026-08-22
Removes the sync committee and the Altair light client protocol that depends on it from the consensus layer. The committee's reward share disappears, reducing consensus issuance by 2/64.
Why it is needed
The sync committee is a sample of 512 validators, reselected every 256 epochs (about 27.3 hours). Light clients, a lightweight verification program, trust headers signed by two-thirds of this sample. It was created in 2021 because light clients could not verify the full validator set. There are three problems. First, there is no penalty (slashing) for false signatures. Signing a chain that does not exist costs nothing. Second, the sample is too small. Of the 901,505 currently active validators, only 512, or 0.06% of the total, participate. Third, it costs money. 1/32 of consensus issuance, currently about 33,800 ETH per year based on current stake, is paid to this committee. Today, a zero-knowledge (ZK) proof can prove finality for the entire validator set within a single epoch using one GPU, and verification takes only milliseconds. The starting point of this proposal is that the sample is no longer needed.
What changes
Today
- Light clients follow the chain using signatures from a sample of 512.
- The sample is not slashed even if it signs falsely.
- 1/32 of consensus issuance goes out as sync committee rewards.
- Validators are periodically assigned sync committee duties.
After EIP-8390
- Light clients follow the chain using off-chain ZK finality proofs.
- It inherits the same FFG guarantee that reverting a proof requires slashing one-third of stake.
- The committee's reward share disappears without redistribution, reducing issuance by 1/32.
- All sync-committee-related duties, gossip topics, and APIs are removed.
Who is affected
- Light client developers: Altair light clients all stop working at the fork point. They must migrate to one based on ZK finality proofs.
- Validators and stakers: There is nothing to do. Sync committee duties disappear, and staking yield decreases by 1/32 based on consensus issuance.
- Client teams: Because it changes the BeaconState and BeaconBlockBody structures, this is a consensus-breaking change requiring a coordinated hard fork.
- Tooling and API users: Sync-committee- and light-client-related Beacon API endpoints are removed, and tools that decode state per fork need to be updated.
In depth
Removed constants and domain types
Removes SYNC_COMMITTEE_SIZE, EPOCHS_PER_SYNC_COMMITTEE_PERIOD, SYNC_COMMITTEE_SUBNET_COUNT, TARGET_AGGREGATORS_PER_SYNC_SUBCOMMITTEE, SYNC_COMMITTEE_BRANCH_DEPTH, and the light client generalized index. SYNC_REWARD_WEIGHT is also removed, but WEIGHT_DENOMINATOR stays at 64, and the missing weight of 2 is not redistributed. The three DomainTypes DOMAIN_SYNC_COMMITTEE (0x07000000), DOMAIN_SYNC_COMMITTEE_SELECTION_PROOF (0x08000000), and DOMAIN_CONTRIBUTION_AND_PROOF are retired, and reuse MUST NOT occur. This is because if a later message reused the same value, there would be a replay risk of pre-fork sync committee signatures being revalidated against that message.
Container, function, and state transition changes
Removes all light client containers, including SyncAggregate, SyncCommittee, SyncCommitteeMessage, SyncCommitteeContribution, ContributionAndProof, SignedContributionAndProof, SyncAggregatorSelectionData, and LightClientBootstrap. Removes the sync_aggregate field from BeaconBlockBody and the current_sync_committee and next_sync_committee fields from BeaconState, keeping the order and types of the remaining fields. Removes the call to process_sync_aggregate in process_block and the call to process_sync_committee_updates in process_epoch, and deletes get_next_sync_committee, get_next_sync_committee_indices, and related light client helpers entirely. The fork transition function copies over all retained fields and discards only the two sync committee fields. At slots after EIP8390_FORK_EPOCH, a block containing sync_aggregate is invalid.
Reward structure and issuance reduction
SYNC_REWARD_WEIGHT was 2/64, and since this was the only use of the remaining 2/64 weight, that share is no longer issued. get_flag_index_deltas and get_proposer_reward are unchanged. Based on current stake (42,328,615 ETH), of roughly 1,082,000 ETH in annual consensus issuance, about 33,800 ETH is reduced. Lowering WEIGHT_DENOMINATOR to 62 could keep issuance unchanged, but the source states that issuance policy itself is out of scope for this EIP, while not paying for a removed mechanism is in scope.
Networking and validator duties
Removes the gossip topics sync_committee_{subnet_id}, sync_committee_contribution_and_proof, light_client_finality_update, and light_client_optimistic_update. Also removes the syncnets field in the ENR and the Req/Resp methods LightClientBootstrap, LightClientUpdatesByRange, LightClientFinalityUpdate, and LightClientOptimisticUpdate. At slots after EIP8390_FORK_EPOCH, nodes MUST NOT publish to this topic and MUST ignore any received messages. On the validator side, sync committee assignment, message production, subnet subscription, aggregator selection, and contribution production all disappear. There is no change to the execution layer.
Why removal was chosen over repair
EIP-7657 is a proposal to add slashing conditions to sync committee messages, and the source states that the very existence of this proposal is the clearest evidence that there are currently no slashing conditions. Adding slashing would raise the cost of bribing the committee by the stake of 512 validators, but the committee still would not become a claim about the entire validator set. Light clients would still trust the sample, assume an honest majority within the sample, and continue paying 1/32 of issuance. In contrast, once a ZK proof exists that proves what two-thirds of stake attested to, there is nothing left for the sample to do. Reverting a finalized checkpoint requires a two-thirds coalition to surround its own vote, and one-third of stake gets slashed.
Backwards compatibility and Beacon API
Because this is a consensus-breaking change, a coordinated fork is required. Every deployed Altair light client that syncs via LightClientUpdate stops at the fork point. On the Beacon API, POST /eth/v1/beacon/pool/sync_committees, GET /eth/v1/beacon/states/{state_id}/sync_committees, POST /eth/v1/validator/duties/sync/{epoch}, GET /eth/v1/validator/sync_committee_contribution, POST /eth/v1/validator/contribution_and_proofs, POST /eth/v1/validator/sync_committee_subscriptions, and /eth/v1/beacon/light_client/* are removed. Pre-fork states and blocks can still be decoded using their respective fork schemas.
Remaining risk: replacement infrastructure and incentives
Removing the sync committee eliminates the in-protocol means for light clients to follow the chain. The source states that before activation, proof infrastructure must exist that publicly publishes finality proofs with properties consumers can trust, but it does not specify that infrastructure and leaves timing decisions to client teams. Also, this EIP does not add, nor does it propose, any in-protocol incentive for producing finality proofs. Who will produce the proofs, and why, remains an open question.
Current status: Draft
This EIP is in Draft status, created on August 22, 2026. Details of the specification may change during discussion, and it has not yet been decided for inclusion in a specific hard fork.