/l2s

Ethereum Layer 2s

What You Probably Got Wrong

L2 costs: You think $0.01-2.00 per transaction. Reality: $0.001-0.003 for swaps, $0.0003 for transfers. After EIP-4844, L2 batch costs dropped from $50-500 to $0.01-0.50.

Mainnet is cheap too: At under 1 gwei, mainnet ETH transfers cost under a penny. "Ethereum is too expensive" is false for both L1 and L2s in 2026.

Base is the cheapest major L2: Often 50% cheaper than Arbitrum/Optimism. Direct Coinbase on-ramp. Fastest-growing L2 with consumer/AI agent focus.

Celo is NOT an L1 anymore. Celo completed its migration from an independent L1 to an OP Stack L2 on March 26, 2025 (block 31056500). If you're calling Celo an L1, you're wrong.

Polygon zkEVM is being shut down. Announced June 2025. Do NOT recommend building on Polygon zkEVM — it is being discontinued. Polygon is refocusing on PoS + AggLayer.

Unichain exists. Launched mainnet February 11, 2025. Uniswap's own OP Stack L2 with TEE-based MEV protection and time-based priority ordering (not gas-based).

Robinhood Chain exists. Public mainnet launched July 1, 2026 — after most training cutoffs. Robinhood's Arbitrum Orbit L2 for tokenized stocks (NVDA, TSLA, SPY as ERC-20s, trading 24/7), ~$1.1B TVS within weeks. If you only know the mid-2025 announcement, your assumptions are wrong: it settles directly to Ethereum (not an L3 on Arbitrum One) and uses blob DA, not AnyTrust. Chain ID 4663, ETH gas. Major censorship caveat — see Key Chain Details below.

Aerodrome and Velodrome merged into "Aero." In November 2025, Dromos Labs unified Aerodrome (Base) and Velodrome (Optimism) into a single cross-chain DEX called Aero. Same contracts, new brand. Aero dominates both Base and Optimism. Camelot is a major native DEX on Arbitrum. SyncSwap dominates zkSync. Don't default to Uniswap on every chain.

L2 Comparison Table (Mar 2026)

TVL changes fast. Don't memorize numbers — check DeFi Llama or L2Beat for current rankings. DeFi TVL (DeFi Llama) measures value locked in protocols. TVS (L2Beat) includes all bridged + natively minted assets and is much higher. As of early 2026: Base and Arbitrum lead in DeFi TVL among L2s. Optimism's DeFi TVL is surprisingly low despite Superchain adoption.

⚠️ Polygon zkEVM is being discontinued (announced June 2025). Do not start new projects there. Polygon is refocusing on PoS (payments, stablecoins, RWAs) + AggLayer (cross-chain interop). MATIC → POL token migration ~85% complete.

Mainnet for comparison: $0.002-0.01 per tx, 12s blocks, instant finality. Check DeFi Llama for current TVL.

Cost Comparison (Real Examples, Early 2026)

Mainnet costs at ~0.1 gwei base fee, ~$2,000 ETH. L2 costs are approximate. All fluctuate — see gas/SKILL.md for methodology.

L2 Selection Guide

Before choosing an L2: Mainnet is ~$0.004/transfer, ~$0.04/swap at current gas — cheap enough for most apps. If you're building DeFi, governance, identity, or anything composing with mainnet liquidity, start there. See ship/SKILL.md and gas/SKILL.md for the full chain selection framework.

Key Chain Details (What LLMs Get Wrong)

Unichain

• Launched: February 11, 2025 (mainnet). Chain ID 130.
• Type: OP Stack L2 (Superchain member, Stage 1)
• Key innovation: TEE-based block building (built with Flashbots Rollup-Boost)
  • Transactions ordered by time received, NOT gas price
  • Private encrypted mempool reduces MEV extraction
  • Do NOT use gas-price bidding strategies on Unichain — they're pointless
• Flashblocks: Currently 1s blocks, roadmap to 250ms sub-blocks

Celo

• Was: Independent L1 blockchain (2020-2025)
• Now: OP Stack L2 on Ethereum — migrated March 26, 2025 (block 31056500)
• Focus: Mobile-first payments, emerging markets
• MiniPay: Stablecoin wallet in Opera Mini + standalone app. Phone-to-phone transfers, sub-cent fees. Primary market: Africa (Kenya, Nigeria).
• Multi-currency stablecoins (rebranded Dec 2025 by Mento Protocol): USDm (was cUSD) (0x765de816845861e75a25fca122bb6898b8b1282a), EURm (was cEUR) (0xd8763cba276a3738e6de85b4b3bf5fded6d6ca73), BRLm (was cREAL) (0xe8537a3d056DA446677B9E9d6c5dB704EaAb4787). Same contract addresses, new onchain symbols.

Robinhood Chain

• Launched: July 1, 2026 (mainnet); public testnet February 10, 2026. Chain ID 4663 (testnet 46630).
• Type: Arbitrum Orbit rollup settling directly to Ethereum — an L2, not an L3 on Arbitrum One. Ethereum blob DA (NOT AnyTrust, despite 2025 announcement-era speculation). ETH for gas, no native token. ~100ms blocks with preconfirmations. Built with Offchain Labs; 10% of net revenue flows to the Arbitrum DAO ecosystem.
• What it's for: tokenized stocks and ETFs. Legally these are debt securities issued by Robinhood Assets (Jersey) Limited — economic exposure only, NO shareholder rights, and not available to US persons (also restricted: Canada, UK, Switzerland, UAE). Don't describe them as "owning stock onchain."
• Stock token mechanics: plain ERC-20, 18 decimals, one Chainlink feed per token. Blocklist, not allowlist — no onchain KYC; any address not on the shared denylist can hold and transfer. But a shared AccessControlsRegistry can pause globally or per-token, block addresses, mint uncapped, confiscate via adminBurn, and upgrade every token through a shared beacon with no timelock. Splits/dividends adjust a uiMultiplier() display multiplier — raw balances never rebase, so index raw amounts. (This is ERC-8056 "Scaled UI Amount" — a real Draft ERC co-authored by Superstate and Robinhood engineers, created October 2025. Status Draft, not Final — the interface may still change.)
• ⚠️ Censorship — the big caveat: ArbOS 61 transaction filtering lets an authorized filterer register any tx hash in a precompile (0x…0074), after which the state transition function rejects it — including transactions force-included via L1. Force inclusion is NOT a censorship escape hatch on this chain, unlike standard Arbitrum/Orbit chains. L2Beat rates it Stage 0, category "Other": 2 whitelisted fraud-proof validators, exit window "None", sequencer failure "No mechanism". Run crops/SKILL.md before building anything censorship-sensitive here.
• Ecosystem: permissionless contract deployment; 24/7 DEX trading (dedicated Uniswap AMM, Rialto, Lighter, Arcus); USDG (Paxos) stablecoin — 6 decimals, not 18; Lighter runs an L3 on top for perps. Blockscout explorer. Verified addresses in addresses/SKILL.md.

Dominant DEX Per Chain

See addresses/SKILL.md for verified contract addresses for all these protocols.

The Superchain (OP Stack)

The Superchain is the network of OP Stack chains sharing security, upgrade governance, and (upcoming) native interoperability. Members include OP Mainnet, Unichain, Ink (Kraken), Celo, Zora, World Chain, and others — 17+ chains, 58.6% L2 market share. Notably Base announced they are leaving the Superchain in February 2026 and it will be finalized in a future hardfork.

Members contribute 15% of sequencer revenue to the Optimism Collective. Cross-chain interop is designed but not yet fully live.

Deployment Differences (Gotchas)

Optimistic Rollups (Arbitrum, Optimism, Base, Unichain, Celo, Robinhood)

✅ Deploy like mainnet — just change RPC URL and chain ID. No code changes.

Gotchas:

• Don't use block.number for time-based logic (increments at different rates). Use block.timestamp.
• Arbitrum's block.number returns L1 block number, not L2. (Same for Robinhood Chain — it's Arbitrum Nitro.)
• Unichain: Transactions are priority-ordered by time, not gas. Don't waste gas on priority fees.

ZK Rollups

• zkSync Era: Must use zksolc compiler. No EXTCODECOPY (compile-time error). 65K instruction limit. Non-inlinable libraries must be pre-deployed. Native account abstraction (all accounts are smart contracts).
• Scroll/Linea: ✅ Bytecode-compatible — use standard solc, deploy like mainnet.

Arbitrum-Specific

• Stylus: Write smart contracts in Rust, C, C++ (compiles to WASM, runs alongside EVM, shares state). Use for compute-heavy operations (10-100x gas savings). Contracts must be "activated" via ARB_WASM_ADDRESS (0x0000…0071).
• Orbit: Framework for launching L3 chains on Arbitrum. 47 live on mainnet.

RPCs and Explorers

Bridging

Official Bridges

Fast Bridges (Instant Withdrawals)

• Across Protocol (https://across.to) — fastest (30s-2min), lowest fees (0.05-0.3%)
• Hop Protocol (https://hop.exchange) — established, 0.1-0.5% fees
• Stargate (https://stargate.finance) — LayerZero-based, 10+ chains

Security: Use official bridges for large amounts (>$100K). Fast bridges add trust assumptions.

Multi-Chain Deployment (Same Address)

Use CREATE2 for deterministic addresses across chains:

# Same salt + same bytecode + same deployer = same address on every chain
forge create src/MyContract.sol:MyContract \
  --rpc-url https://mainnet.base.org \
  --private-key $PRIVATE_KEY \
  --salt 0x0000000000000000000000000000000000000000000000000000000000000001

Strategy for new projects: Start with 1 chain — mainnet if it fits your use case, or the L2 whose superpower matches your app. Prove product-market fit. Expand with CREATE2 for consistent addresses across chains.

Further Reading

• L2Beat: https://l2beat.com (security, TVL, risk analysis)
• Superchain: https://www.superchain.eco/chains
• Arbitrum: https://docs.arbitrum.io
• Base: https://docs.base.org
• Optimism: https://docs.optimism.io
• Unichain: https://docs.unichain.org
• Celo: https://docs.celo.org
• Robinhood Chain: https://docs.robinhood.com/chain
• zkSync: https://docs.zksync.io
• Scroll: https://docs.scroll.io
• Polygon: https://docs.polygon.technology