January 2026 was supposed to be a milestone moment for Layer 2 scaling. Arbitrum One, OP Mainnet, and Base all achieved Stage 1 decentralization under L2BEAT’s classification framework. Arbitrum’s BoLD (Bounded Liquidity Delay) went live, enabling anyone to submit permissionless fraud proofs without whitelisting. This was years in the making—a genuine technical achievement.
But here’s the uncomfortable truth I’ve been wrestling with as someone who’s spent 6 years building L2 infrastructure: I think we’re confusing progress with actual decentralization.
What Stage 1 Actually Means (And Doesn’t)
Let’s be precise about what we achieved:
Permissionless fraud proofs: Anyone can challenge invalid state transitions
Trustless exits: Users can withdraw funds even if the sequencer goes rogue
Improved security model: Fraud proofs prevent fund theft
This is security, and it matters enormously. But it’s NOT liveness or censorship resistance.
Here’s what we DIDN’T achieve:
Decentralized transaction ordering
Censorship resistance at the sequencer level
Multiple independent operators
The Sequencer Problem Nobody’s Talking About
Despite Stage 1, the sequencer architecture hasn’t changed:
- Arbitrum: Offchain Labs/Arbitrum Foundation controls the single sequencer
- Base: Coinbase is the sole sequencer operator
- OP Mainnet: Optimism Foundation runs the sequencer
These entities control transaction ordering, which means they can:
- Censor specific transactions or addresses
- Reorder transactions for MEV extraction
- Comply with government sanctions (like OFAC lists)
Fraud proofs only catch misbehavior AFTER it happens. They prevent the sequencer from stealing your funds, but they don’t prevent censorship in real-time.
Comparison: L2 Sequencers vs. Ethereum Validators
Let’s contrast this with Ethereum mainnet:
Ethereum L1:
- Anyone with 32 ETH can run a validator (permissionless)
- ~1 million validator nodes globally
- No single entity controls transaction inclusion
L2 Sequencers (Stage 1):
- Single operator controls sequencer
- Permissioned—you can’t just spin up a competing sequencer
- Centralized transaction ordering
We’ve built a beautifully decentralized proof system (fraud proofs) on top of a centralized transaction ordering layer. That’s not rebranding… it’s an architectural reality we need to acknowledge.
What Happened to Shared Sequencers?
There ARE solutions in development:
Espresso Systems: Building shared sequencer network using HotShot consensus + EigenLayer for security. VM-agnostic, supports zkVMs and optimistic rollups. Still early-stage.
Astria: Was building shared sequencer using CometBFT (Tendermint fork). Shut down entirely in 2025, highlighting how hard the coordination problem is.
The market is speaking: sequencer decentralization is not being prioritized relative to other roadmap items. Why?
Did We Just Rebrand Centralization?
Here’s my controversial take: calling these networks “Stage 1 decentralized” feels like a rebranding exercise.
Yes, L2BEAT’s framework is transparent about what Stage 1 means. Yes, fraud proofs are a massive technical achievement. Yes, roadmaps always included sequencer decentralization as a future goal.
But when we tell users “Arbitrum is decentralized now,” what do they hear? Do they understand that fraud proofs protect fund security but not censorship resistance? Or do they assume the network is now permissionless and censorship-resistant like Ethereum L1?
The first real test will come when Coinbase receives a legal order to censor specific addresses on Base. Will they comply? (Probably yes—they’re a publicly traded US company.) Will users even know their transactions were censored? Will forced inclusion via L1 be practical for average users?
What Actually Matters?
I keep coming back to this question: What do we actually care about?
If we care about fund security: Stage 1 is a huge win. Fraud proofs work.
If we care about censorship resistance: We haven’t solved this yet, and the market doesn’t seem that concerned.
If we care about credible neutrality: Single-operator sequencers are not neutral by definition.
I’m not saying Stage 1 is meaningless. It’s a critical foundation. But let’s be honest about what we’ve achieved and what we haven’t.
The question for this community: Do you think sequencer centralization matters? Or is fund security (via fraud proofs) enough? At what point does “Stage 1 with centralized sequencer” become misleading?
I’d love to hear from builders, users, and researchers on this.
Sources:
- Arbitrum’s 2026 Roadmap (BlockEden.xyz)
- Layer 2 Consolidation War (BlockEden.xyz)
- Arbitrum One (L2BEAT)
- 2026 Layer 2 Outlook (The Block)
- Deep Dive into Layer 2 Sequencers (Orochi Network)