Last Tuesday, a project I had been tracking since its seed round announced a $100 million ecosystem fund. The press release was immaculate. The accompanying thread—complete with the word "decentralization" repeated eleven times in nine posts—was already circulating through every group chat I am in. But buried in the seventh paragraph of the technical documentation, past the governance diagram and the tokenomics pie chart, was a single line that stopped me cold: the sequencer operator, at launch, would be "a permissioned entity initially transitioning toward a multi-operator model." The word "initially" had no date attached. I have seen this sentence before. I have seen it in the documentation of nearly every major Layer 2 that has raised capital in the past three years. And every time, it means the same thing: the thing being sold as infrastructure is, for all practical purposes, a single server run by a company that answers to a board of directors.
This is the paradox of transparency in a cashless society, translated into the grammar of rollups. The chain is transparent—you can watch every transaction settle to Ethereum. What is opaque is the machine that orders those transactions, decides which ones to include, and can, in theory, reorder them for its own profit. The transparency of the ledger disguises the opacity of the operator. We celebrate the visibility of the output and ignore the invisibility of the input. The sequencer is the most powerful single point of failure in the modern L2 stack, and almost no one is talking about who controls it because the conversation has been captured by a governance token that does not actually govern the sequencer.
I want to walk through the architecture of this specific launch, because it is a useful case study, but also because the pattern is now so uniform that you could swap the names and publish the same analysis about a dozen other projects. I am going to focus on the technical specifics—the operational reality beneath the marketing—and then I want to make an argument that runs against the current consensus: that the L2 decentralization roadmap is not delayed, it is structurally incapable of arriving in the form currently promised, and that this is not an engineering problem but an incentive problem. The people who control the sequencer have no financial reason to give it up, and the governance token they are distributing is designed to make you believe they have.
Let me start with what the documentation actually says, because the details matter and the marketing obscures them. The project in question—let me call it a "leading optimistic rollup," which is the industry's way of saying "an optimistic rollup with a sequencer everyone forgets to ask about"—describes three phases of decentralization. Phase one, "training wheels," is the current state: a centralized sequencer, a multisig upgrade key held by the founding team, and a fraud proof system that is, in the documentation's own words, "under active development." Phase two, "community governance," involves a token launch and a DAO that will "oversee" the sequencer transition. Phase three, "permissionless operation," is described as a "long-term objective" with no timeline. The document is nine pages long. Pages one through six are about throughput and gas fees. Page seven is the decentralization roadmap. Pages eight and nine are the token allocation.
The ratio is the message. The economic design gets two pages of detail; the control structure gets one paragraph of aspiration. And this is not an outlier—I have read the documentation for every major L2 in the top twenty by TVL, and the pattern holds almost without exception: the more a project talks about speed and cheap transactions, the less it says about who runs the sequencer and what happens when that entity fails, censors, or sells.
Here is the technical reality that the documentation elides. In an optimistic rollup, the sequencer is the component that receives transactions from users, orders them, and submits them to Ethereum as a batch. It does not need to be trusted for correctness—that is what fraud proofs are for, eventually. But it absolutely must be trusted for liveness and censorship resistance, and in the current architecture, it is also the entity that captures MEV, which is a polite way of saying it is the entity that profits from reordering your transactions to its own advantage. The fraud proof system, if it works and if it is actually deployed and if there is a mechanism to force a withdrawal through an alternative path when the sequencer goes down, protects the correctness of the state. It does not protect your ability to transact. It does not protect you from being front-run by the operator. And it certainly does not protect you from the operator deciding, on a Tuesday, that a certain category of transactions is no longer welcome.
I spent three weeks in early 2024 mapping sequencer configurations across the major L2s. I built a spreadsheet—I am an analyst, I build spreadsheets—that tracked, for each project: the number of sequencer operators, the upgrade key structure, the fraud proof status, the forced withdrawal mechanism, and the nominal annualized yield on the governance token. The correlation was striking. The projects with the highest yields and the most aggressive incentive programs were, without exception, the ones with the most centralized sequencer control. The projects with genuine multi-operator setups—there were two, and I am not going to name them because I do not want this to become a hit piece against the good actors—had lower yields and smaller ecosystem funds. Decentralization does not pay for marketing. Marketing pays for marketing, and centralized sequencers are extremely profitable for the entities that run them.
This is where my skepticism about "decentralized sequencing" hardens into something closer to resignation. I have watched this movie for two years now. Each cycle produces a new cohort of projects that promise the sequencer will be decentralized "soon." Each cycle, the same projects launch a governance token, distribute it to early users, and then discover that the token holders have no actual authority over the sequencer—the governance proposals are binding only if the founding team chooses to implement them, and the founding team, as the operator of the sequencer, has every incentive to stall. The token becomes a voting mechanism for parameters that do not matter, while the parameters that do matter—who orders transactions, who captures MEV, who can pause the chain—remain with the multisig.
And the market, in its current euphoric state, does not care. This is the part that keeps me up at night. In a bull market, nobody asks who runs the sequencer because everyone is making money. The token price goes up, the TVL goes up, the ecosystem fund attracts developers who are paid in the token, and the question of control becomes an abstraction that you can safely ignore as long as the yields are high. But the mechanics of the sequencer do not change based on the price of the token. The centralization is a latent structural risk that only becomes visible during a stress event—a regulatory action against the operator, a hack of the operator's infrastructure, a deliberate censorship decision, or simply a period of congestion during which the operator prioritizes its own MEV extraction over user transactions.
I documented this dynamic in 2020, during the first DeFi summer, and I am watching it repeat with a new set of actors. The protocols I audited then were promising revolutionary financial inclusion while their tokenomics were designed to extract value from the exact populations they claimed to serve. The L2s I am auditing now are promising a decentralized future while their sequencer architecture concentrates control in a way that would make a traditional payment processor blush. The vocabulary has changed. The pattern has not.
Let me be precise about what I am not arguing. I am not arguing that optimistic rollups are a failure. Fraud proofs, when they work, are a genuine innovation, and the settlement guarantees to Ethereum are real. I am not arguing that the teams building these projects are malicious—most of them are sincere, and I have had productive conversations with several. What I am arguing is that the decentralization rhetoric has become untethered from the decentralization roadmap, and that the gap between the two is not an engineering accident but a structural feature. The people who control the sequencer have a profit motive to keep controlling it. The governance token is designed to give the appearance of distributed control without the substance. And the market, in its current state, rewards the appearance.
There is a contrarian reading of this, and I want to take it seriously before I dismiss it. The argument goes like this: centralized sequencing is actually fine, because the sequencer submits batches to Ethereum, and Ethereum provides the ultimate guarantee. If the sequencer misbehaves, users can exit to L1, and the fraud proof system will eventually adjudicate disputes. The decentralization roadmap is a transition to a more robust operational model, not a fundamental change in security guarantees, and the market is correct to price in the settlement layer's security while discounting the sequencer's centralization.
I think this argument is wrong, and I want to explain why in a way that goes beyond the obvious "forced withdrawal is hard." The deeper problem is that the exit option is not available to the populations that L2s claim to serve. The unbanked user in Lagos or Jakarta who is using an L2 because gas fees on L1 are prohibitive cannot afford to exit to L1 during a censorship event—the costs of exiting are precisely what the L2 was designed to reduce. The security guarantee is theoretical for anyone who lacks the capital to exercise it. This is the silence between transactions that I have been listening to for years: the people for whom the rollup is a necessity, not a choice, and who have no way to verify the claims being made about the sequencer's eventual decentralization. The exit option is a right, but like all rights, it is only meaningful if you can afford to exercise it.
And the governance token, the mechanism that is supposed to bridge this gap, is structurally incapable of doing so. The token holders can vote on whether to fund a new developer incentive program. They cannot vote on whether the operator should be replaced. The operator is a legal entity with a contract, and the token holders are a dispersed group of retail investors with no legal standing. The governance proposals that matter—the ones that would actually decentralize the sequencer—require the operator to voluntarily give up its position, and the operator's revenue depends on it not doing so. This is not a governance failure. It is a governance design. The token exists to absorb the pressure for decentralization without actually delivering it.
I have been thinking about this in the context of the AI-driven forecasting work I have been doing with a small team over the past year. We have been building models that analyze on-chain liquidity data against macro indicators, and one of the things that has become clear is that the centralization of sequencing is a hidden variable in liquidity risk. When a sequencer goes down—and they do go down, with a regularity that the marketing materials never mention—the liquidity on that rollup does not simply pause; it evaporates to alternative venues, and the users who are least able to move their capital are the ones who bear the cost. The models did not predict this because the models do not see the sequencer; they see the TVL, and the TVL is a lagging indicator. The real risk is in the operational architecture, and that architecture is deliberately opaque.
So where does this leave us? I am not going to end with a list of recommendations, because the problem is structural and a checklist is a way of avoiding the structural argument. What I will say is this: the next time a project announces a $100 million ecosystem fund and a nine-page document with one paragraph on decentralization, read that paragraph first. Read it twice. Ask what the word "initially" means, and ask who benefits from it remaining undefined. The answer to that question is the most important piece of information in the entire announcement, and it is almost never the one the project wants you to read.
The cycle we are in rewards speed and cheap transactions and the appearance of decentralization. The next cycle, whenever it arrives, will reward the projects that actually delivered it, or it will reward the projects that survived the failure of the ones that did not. The sequencer is the ghost in the machine, and the machine is running on a centralized server.
I spent six months in 2017 tracking the relationship between the naira's devaluation and Bitcoin wallet creation in Lagos, and the lesson I took from that research was that adoption driven by necessity is more durable than adoption driven by speculation. The same principle applies here, but in reverse. The adoption of L2s is being driven by speculation and subsidized yields, and the mechanism that makes them work—the sequencer—is being managed by entities whose interests are aligned with the subsidy, not with the durability. When the subsidy ends, the sequencer will still be there, and so will the question of who controls it. The only thing that will have changed is who can afford to ask.