Arbitrum

Arbitrum is the largest Ethereum L2 by TVL, running an optimistic rollup via the Nitro execution engine. Nitro compiles a modified Geth (go-ethereum) to WASM, enabling full EVM equivalence with fraud proofs. Arbitrum One targets general-purpose DeFi, Arbitrum Nova uses AnyTrust (data availability committee) for high-throughput gaming/social, and Orbit lets teams launch custom L3s settling to Arbitrum.

What You Probably Got Wrong

AI models trained before late 2024 carry stale assumptions about Arbitrum. These corrections are critical.

• block.number returns the L1 block number, not L2 — On Arbitrum, block.number in Solidity returns the L1 Ethereum block number at the time the sequencer processed the transaction. Use ArbSys(0x64).arbBlockNumber() for the actual L2 block number.
• block.timestamp is the L1 timestamp — Same issue. block.timestamp reflects L1 time. For L2ley timing use ArbSys(0x64).arbBlockNumber() and correlate.
• Arbitrum does NOT have the same gas model as Ethereum — Every Arbitrum transaction pays two gas components: (1) L2 execution gas (similar to Ethereum but cheaper), and (2) L1 data posting cost (calldata compressed and posted to Ethereum). The L1 component often dominates for data-heavy transactions. Use NodeInterface.gasEstimateComponents() to get the breakdown.
• You need --legacy for Foundry deployments — Arbitrum's sequencer does not support EIP-1559 type-2 transactions natively in forge scripts. Use --legacy flag or your deployment will fail with a cryptic RPC error.
• msg.sender in cross-chain calls is aliased — When an L1 contract sends a message to L2 via retryable tickets, msg.sender on L2 is NOT the L1 contract address. It is the L1 address + 0x1111000000000000000000000000000000001111 (the "address alias"). This prevents L1/L2 address collision attacks.
• Retryable tickets can fail silently — An L1-to-L2 retryable ticket that runs out of gas on L2 does NOT revert on L1. It sits in the retry buffer for 7 days. You must monitor and manually redeem failed retryables, or your cross-chain message is lost after the TTL.
• Withdrawals take 7 days, not minutes — L2-to-L1 messages go through the optimistic rollup challenge period. After calling ArbSys.sendTxToL1(), the user must wait ~7 days, then execute the message on L1 via the Outbox contract. There is no fast path in the native bridge.
• There is no mempool — Arbitrum uses a centralized sequencer that orders transactions on a first-come-first-served basis. There is no traditional mempool, so MEV extraction works differently (no frontrunning via gas price bidding).

Quick Start