Silence is the first vote in a true consensus.
I was sitting in a cafe in Tallinn, reading the post-mortem of yet another Layer 2 outage. The sequencer had stopped producing blocks for 47 minutes. The team apologized, blamed a software upgrade, and promised to decentralize sequencer selection next quarter. I have heard that promise a hundred times. The same team, three years prior, had raised a $50 million Series A on the explicit claim that their rollup would be fully decentralized from day one. The architecture of trust is built on scrutiny, not belief.
This is not a lone incident. In the past six months, I have audited the governance structures of eight major L2 projects. I found that seven of them retain a single sequencer controlled by a multi-sig wallet with 3-of-5 signers—all employees of the founding company. One project had a 2-of-2 multi-sig with one key held by the CEO and the other by a venture partner from the lead investor. Decentralization is a practice, not a feature; yet we continue to buy the narrative that these systems are trustless.
Let me rewind to the summer of 2017. I was a senior researcher in a Tallinn-based cybersecurity firm, and we were asked to conduct a post-mortem of The DAO hack. I spent four months auditing Etherscan transaction logs, identifying 14 critical logical flaws in the reentrancy vulnerability. What struck me most was not the code failure, but the moral vacuum in the smart contract design. The developers were technically brilliant, but they had never considered the ethical implications of a race condition that could drain a DAO of its entire treasury. I wrote a 30-page whitepaper titled "Code is Not Law: The Moral Vacuum in Smart Contracts." It circulated among early adopters, many of whom were disillusioned by the profit-driven narrative. That experience embedded in me a habit—a compulsion, really—to audit not just the code, but the governance architecture surrounding that code.
Fast-forward to today’s bull market. Euphoria masks technical flaws. Every week a new L2 launches with a polished website, a celebrity advisor, and a token that doubles in price within hours. But when I look under the hood, I see centralized sequencers, upgradeable contracts with admin keys, and governance models that concentrate voting power in the hands of the founding team. The architecture of trust is built on scrutiny, not belief. So let me scrutinize.
The Centralization That Silences the Network
The fundamental promise of a rollup is that it inherits the security of Ethereum. Transactions are batched off-chain, compressed into a proof, and submitted to L1. The sequencer, which orders those transactions, is supposed to be permissionless—anyone should be able to submit a batch. In practice, however, almost every L2 uses a single, private sequencer. This sequencer has total control over transaction ordering, which means it can extract MEV at will, censor addresses, and reorder transactions to maximize its own profit. The community has no recourse except to trust the sequencer operator.
I have seen data from Dune Analytics that shows for the top five L2s by TVL, over 98% of transactions are processed by their respective official sequencers. The remaining 2% come from forced inclusion on L1—a mechanism that is both expensive and slow. In effect, users have no meaningful alternative. They are using a service that looks like a decentralized blockchain on the surface, but operates like a centralized database beneath. The architecture of trust is built on scrutiny, not belief.
During my six-week retreat on Estonia’s Hiiumaa island in 2022, after the collapse of FTX, I disconnected from all social media and reviewed my past five years of work. I realized that much of what we call “innovation” in crypto is simply financial engineering disguised as progress. The L2 narrative is a perfect example. We have replaced a slower, more expensive, but truly decentralized execution layer (Ethereum L1) with a fast, cheap, but centralized execution layer (L2). The trade-off is rarely disclosed to retail users. They see low fees and high throughput, but they do not see the single point of failure in the sequencer. Decentralization is a practice, not a feature.
The ZK Proof Cost: An Elephant in the Room
But let us go deeper. Even if sequencers were distributed, the cost of generating zero-knowledge proofs for each batch remains exorbitantly high. I have consulted with several L2 teams on their proving system economics. For a typical ZK rollup, the cost to generate a proof for a batch of 1,000 transactions can range from $500 to $2,000, depending on the circuit complexity. At current gas prices (roughly 10 gwei on Ethereum), the cost of submitting that proof to L1 is about $30. The proving cost dominates. This means that only well-capitalized entities can run provers. The dream of a fully decentralized prover network—where any participant can generate and submit proofs—remains economically unviable until gas returns to bull-market levels or proof aggregation advances dramatically.
I recall a closed-door panel in Geneva, in 2024, after the Spot Bitcoin ETF approvals. I had prepared a 20-slide deck titled "Beyond Speculation: Blockchain as a Trust Layer." During the Q&A, an institutional investor asked me why they should invest in any L2 token. I answered honestly: because the market is pricing in a future of full decentralization that does not yet exist. If you are investing in the token, you are betting that the team will eventually relinquish control of the sequencer and the prover. But the incentives are misaligned. The founding team earns transaction fees from the sequencer—why would they give it up? The architecture of trust is built on scrutiny, not belief.
The Oracle Problem: A Decentralized Joke
Now broaden the lens to DeFi. The entire ecosystem relies on oracles to bring off-chain data on-chain. DeFi’s Achilles’ heel is oracle feed latency. Every liquidation event, every flash loan attack, is predicated on a price discrepancy between what the oracle says and the true market price. Chainlink solved the centralization problem by creating a network of node operators, but those nodes are selected by the Chainlink team. The nodes are mostly overlapping and can be pressured to censor data. In my view, Chainlink solving decentralization with centralized nodes is itself a joke. We have simply moved the trust from a single party to a small consortium.
I remember designing a weighted oracle aggregation system for a DeFi lending protocol in 2020. We used three oracle providers, each with their own set of nodes. The protocol was exploited within three weeks because two of the three providers reported the same stale price. The odds of all three nodes failing independently were astronomically low, but they were not independent—they all pulled from the same centralized exchange API. Decentralization is a practice, not a feature.
Bitcoin: From Peer-to-Peer Cash to Wall Street’s Toy
Let us not forget Bitcoin. Post-ETF approval, Bitcoin has become Wall Street’s toy. Satoshi’s vision of “peer-to-peer electronic cash” is dead. The ETF structure requires custody by centralized entities like Coinbase and Fidelity. The coins are not moving. The Lightning Network, touted as the scaling solution, has stagnated at around 5,000 BTC capacity—a fraction of the total supply. The narrative has shifted from “I am my own bank” to “I can buy a Bitcoin ETF in my retirement account.” The architecture of trust is built on scrutiny, not belief.
I spent a week in early 2025 interviewing 20 Bitcoin ETF holders. None of them understood the concept of self-custody. They believed the ETF was a safer way to own Bitcoin because “the SEC regulates it.” They are not wrong about the safety from theft, but they have ceded the philosophical core of Bitcoin. The network still mines blocks in a decentralized manner, but the economic power is concentrated among large holders and institutional custodians. The censorship resistance that made Bitcoin revolutionary is now irrelevant for the majority of its market cap. Decentralization is a practice, not a feature.
A Contrarian Angle: Pragmatism Over Purity
But I must stop myself from falling into pure cynicism. There is a contrarian angle. Perhaps the market is rationally pricing the transition period. Every technology goes through an incubation phase where centralization is a necessary evil. The internet itself started as a government-research network. The first email servers were controlled by a handful of universities. Over time, the infrastructure became more open. The same could happen with L2s: sequencers may eventually be distributed via staking or auctions, and proving costs may drop with advances in hardware acceleration.
In 2020, when I consulted for MakerDAO on governance redesign, we proposed a quadratic voting mechanism. The community was skeptical at first, but after three months of simulation and town halls, the adoption increased unique voters by 40%. That taught me that inclusive governance design works when the community is given the tools and time to learn. We cannot skip the learning phase. The problem is that the current bull market rewards speed over deliberation. Projects that rush to launch with a centralized sequencer get the liquidity and the attention, while slow-moving projects that prioritize decentralization often die in obscurity.
I wrestle with this tension every day. I have designed a decentralized identity protocol for AI agents—ZK-proofs integrated into wallets to prove provenance. The technology works, but it requires a minimum level of trust in the infrastructure. We are not at the point of full autonomous consensus. The architecture of trust is built on scrutiny, not belief, but we must also be willing to extend provisional trust to projects that are transparent about their centralization and have a credible roadmap to decentralization.
Takeaway: Demand Verifiable Decentralization
So what is the actionable takeaway for a reader in this bull market? Stop trusting project roadmaps. Demand verifiable decentralization metrics. Ask your favorite L2: How many sequencers are actively producing blocks? Can I run a sequencer with 32 ETH? What is the penalty for misbehavior? If the team cannot answer these questions with concrete data, the project is not decentralized. It is a centralized database with a token.
I have created an open-source template—a Governance Decentralization Scorecard—that anyone can use to evaluate a protocol. It includes 15 metrics: number of independent sequencers, proportion of nodes run by the founding team, admin key usage in the last 90 days, and so on. I published it on my GitHub last month, and it has been used to evaluate three L2s already. The results are sobering.
Silence is the first vote in a true consensus. The industry is silent about these centralization issues because the money is good. But as an evangelist for decentralization, I believe that ethical clarity will eventually win. We are still early. The bear market will come again, and when it does, the projects that survive will be those that built genuine, verifiable decentralization—not just the narrative.
Consensus requires patience, not speed. We need to slow down, audit the governance, and demand transparency. Until then, keep your coins on L1, and treat every L2 as a bank that you are willing to trust only until you can verify. The architecture of trust is built on scrutiny, not belief.
