A friend sent me a link last week. It was a tweet from a major Layer-2 project celebrating its “fully decentralized sequencer” upgrade. The community cheered. The token pumped. I opened the smart contract on Etherscan and saw it: the sequencer address was still controlled by a single admin key behind a 2-of-3 multisig. The upgrade hadn't changed the fundamental power structure. We didn't do the work we claimed to have done. And I knew I had to write this.
Truth in blockchain isn't just about consensus algorithms or zk-proofs. It's about where the real power to order transactions sits. After eight years in this ecosystem—from manually auditing ICO genesis blocks in 2017 to reverse-engineering a rug-pulled yield farm in 2020—I've learned that technical architecture reveals what marketing narratives hide. Today, I want to take you through the current state of L2 sequencing, why the “decentralized sequencer” claim is mostly a PowerPoint slide, and what we need to actually fix this.
The Context: What Is a Sequencer and Why Should You Care?
Every Layer-2 network—whether it's an optimistic rollup like Arbitrum or a zk-rollup like zkSync—needs someone to decide the order of transactions. That's the sequencer. In a centralized setup, a single entity collects user transactions, orders them, and submits batches to Ethereum. This gives that entity incredible power: it can censor, front-run, reorder, or simply halt the chain. The industry's promise has always been that sequencers would become decentralized—multiple independent parties sequencing collaboratively, eliminating any single point of control.
But the reality is stark. As of mid-2025, the vast majority of L2s still operate with a single sequencer. The “decentralized sequencing” upgrades we've seen often amount to adding a fallback mechanism or a permissioned set of validators, not a truly trustless, permissionless sequencer network. I've spent the past three months auditing the on-chain data for the top ten L2s by total value locked. The results are uncomfortable.
The Core: What the Code Actually Reveals
Let’s start with Arbitrum. Its current sequencer is operated by Offchain Labs. They've proposed a transition to a “decentralized sequencer” called BOLD, but the code for that upgrade shows it still relies on a single sequencer for most operations—the decentralized fallback only kicks in when that sequencer goes offline for a prolonged period. In practice, that means 99.9% of transactions are still ordered by a single entity. The system is not trustless; it's a backup mechanism.
Optimism launched its “OP Stack” with a modular design that claims to allow custom sequencers. But look at the core smart contract: the sequencer variable is owned by a single address. There's a function to change it, but that's controlled by a multisig. During my audit of their bridge contracts in 2023, I found that the sequencer was the only one allowed to submit transaction batches to Ethereum. If that sequencer goes down or turns malicious, the chain stops or becomes completely vulnerable to reorgs. The so-called decentralization of sequencers is a governance promise, not a technical guarantee.
zkSync Era uses a slightly different approach—its sequencer is based on a “validator” set that can propose blocks. But that validator set is permissioned and currently has only three members, all affiliated with Matter Labs. On-chain data shows that 100% of the blocks in the past year were proposed by the same single node. The other validators are dormant or not used. This is not a decentralized sequencer; it's a rotating oligarchy with unanimity requirement.
I could go through Base, Linea, Scroll—the pattern repeats. Every L2 lists “decentralized sequencing” on its roadmap. But when you dig into the actual smart contracts, you find either a single sequencer address, a small multisig, or a permissioned set with no real rotation. The bull market euphoria has allowed these projects to mask this critical flaw with marketing jargon like “progressive decentralization” or “phased rollout.” But the phase never seems to finish.
The Contrarian: Maybe the Market Doesn't Care, But It Should
A common counterargument I hear is: “Centralized sequencing is faster and cheaper. Users don't care about trustlessness as long as the UX is good.” There's truth to that. In developing countries where I see crypto payments being used for survival—due to hyperinflation, not ideology—people don't ask about sequencer decentralization. They ask if the transaction confirms in seconds and costs less than a cent. The industry has prioritized speed over resilience.
But here's the blind spot: the very thing that makes L2 fast—a single sequencer—also makes it a risk. If that sequencer is compromised or pressured by regulators, it can censor transactions, freeze the entire network, or even manipulate balances by reordering. We saw a glimpse of this in 2023 when a major L2 sequencer went down for hours due to a software bug. If that had been a malicious attack, the consequences would have been devastating. The market is pricing these networks based on their total value locked, not on their actual security model. That's a dangerous mismatch.
Moreover, the narrative of “decentralized” L2s is what attracts capital and trust. If the promise is broken, the reputational damage could spark a flight to security. During my yield farming mishap in 2020, I learned that I could not rely on marketing. I had to read the contracts myself. That experience taught me that the gap between what projects say and what they do is where real risk lives.
The Takeaway: Honest Architecture Is the Only Path Forward
We need to stop pretending that a single sequencer with a scheduled upgrade path is a decentralized system. Decentralized sequencing requires a set of independent, permissionless nodes that collectively produce blocks without relying on a single leader. Projects like Espresso and Radius are building this infrastructure, but they have not yet been adopted by any major L2 in production. The technology is still in experimentation.
For now, the honest framing is this: most L2s are not trustless; they are “speed-optimized” networks with reduced trust assumptions compared to L1, but still reliant on a small group of operators. If you are using these networks for high-value DeFi or governance, you need to accept this risk. If you are building on them, you need to demand transparency about the sequencer architecture.
We didn't build this industry to replicate the centralized systems we left behind. Truth in blockchain isn't about being ideologically pure; it's about being intellectually honest. And until we confront the centralization in L2 sequencing, we are building castles on sand.
The next time you see a tweet celebrating a “fully decentralized sequencer,” open the contract. Look at the owner variable. Look at the multisig. Then ask yourself: who really orders the blocks? The answer might surprise you.