Scaling & Layer 2

optimistic rollup

An optimistic rollup is a Layer-2 chain that executes transactions off the main chain but posts their compressed data to it, and is 'optimistic' because it assumes every batch is valid by default rather than proving it. The name captures the trust model: like a shopkeeper who lets regulars run a tab and only checks the receipts if someone disputes a charge, the rollup publishes a new state root and starts treating it as correct immediately, relying on the threat of a fraud proof to deter cheating. Arbitrum and Optimism (OP Stack) are the leading examples and together hold most of Ethereum's L2 activity.

The data is the security. The rollup posts the transaction data for every batch to Ethereum (today as blobs via EIP-4844), so anyone can independently re-execute the batch and recompute what the correct state root should be. If a sequencer posts a state root that does not match honest re-execution, a verifier submits a fraud proof; modern systems use an interactive bisection game that narrows the disagreement down to a single disputed instruction, which Ethereum then checks cheaply on-chain. Because only one honest verifier is needed to catch fraud, the system is secure even if everyone else is malicious — provided the data was published.

The cost of optimism is the challenge period. Since a batch is only presumed valid, withdrawals from the rollup to Ethereum must wait out a dispute window — typically about seven days on Arbitrum and Optimism — long enough for an honest party to detect and prove fraud before funds leave. This makes native withdrawals slow (third-party liquidity bridges paper over it for a fee). The compensating advantage over ZK rollups is engineering simplicity and full EVM equivalence: because there is no need to prove execution in a circuit, optimistic rollups ran general Solidity contracts years before zkEVMs matured.

A sequencer posts batch #1000 claiming the new state root is S. A verifier re-executes the batch from the published data and gets S'. They open a dispute; the bisection game pinpoints the exact opcode where S and S' diverge, Ethereum executes just that one step, finds the sequencer wrong, rejects the batch and slashes the sequencer's bond.

Optimistic rollups are not insecure because they 'don't check' — they are checkable. Security holds as long as at least one honest party watches, can access the published data, and can post a fraud proof before the challenge period ends. Censoring or starving that party of data is the real threat model.