I spent a Saturday in March reading the OP Stack derivation pipeline for the third time. Not hunting a bug this time โ hunting a decision. Two years ago I would have told you the interesting part of a rollup was the proof system. Now I think the interesting part is the paragraph in the governance charter that tells you who can pause the bridge.
Ethereum's Layer 2 competition gets covered as a cryptography race: optimistic fraud proofs versus validity proofs, seven-day withdrawal windows versus minutes, blob capacity versus calldata. That framing is comfortable because it is measurable, and measurability is the cheapest form of authority in this industry. It is also, increasingly, a category error. When the marginal cost of launching a new chain falls close to zero, the binding constraint stops being cryptography and becomes coordination โ who shares a bridge, who shares a sequencer, who shares a veto. The technical artifacts do not disappear. They get demoted to packaging.
Code speaks, but culture listens. In the rollup business, the culture is written down in charters that almost nobody reads, and the packaging is a landing page that everyone does.
Three things you have to hold at once
A rollup stack is not a blockchain. It is a contract suite on Ethereum plus an off-chain software bundle. On the L1 side you get a bridge contract that custodies deposits, an assertion or state-root contract that records what the rollup believes, a verifier if the design is validity-based, and a configuration contract that holds the addresses allowed to change the other three. On the off-chain side you get a sequencer that orders transactions, a batcher that compresses them to the L1, a proposer that submits state commitments, and a derivation path that lets anyone reconstruct the chain from L1 data if the operators vanish.
Deploying a chain means deploying that contract suite and running that software. That is the entire product. It is a deployment, not an invention.
Between 2023 and 2024, four organizations open-sourced that deployment and started giving it away. Optimism released the OP Stack under permissive terms and built a membership brand around it called the Superchain. Offchain Labs released Arbitrum Orbit and let chains pick their own data availability layer. Matter Labs released the ZK Stack and called the result elastic chains. Polygon released the CDK and stitched the output into something called the AggLayer. The marketing was different. The pitch was identical: run your own chain, inherit Ethereum's security, keep your own fee market and your own token.
Then Dencun activated in March 2024 and EIP-4844 introduced blobs โ a separate, cheaper data channel with a target of three blobs per block and a maximum of six. Data costs for rollups fell by roughly an order of magnitude on the chains that adopted them quickly. That single change did two things that nobody fully priced.
First, it made small rollups economically viable. A chain that previously needed meaningful throughput to justify its L1 data bill could now operate on a modest fee base.
Second, and more important, it turned cheap transactions into a commodity. When every chain in a family is cheap, cheapness stops being a reason to exist. Differentiation migrates upward, from the fee to the federation.
The WordPress-ification problem
I keep coming back to an analogy that annoys people. In 2003, the marginal cost of publishing a website collapsed. The winner of that collapse was not the best content management system. It was the one with the largest plugin surface, the largest theme market, the most hosting partners, and the strongest shared mental model of what a website even is. WordPress did not win on code quality. It won on the API surface and the vocabulary.
Rollup stacks are running the same play, and the tell is the word the industry uses: stack. A stack implies layers you can swap. What the vendors actually want is a surface you cannot leave.
The distinction that matters is not which proof system is faster. It is built-with versus part-of.
A chain built with Arbitrum Orbit is, in the ordinary sense, a customer. It runs Offchain Labs software, it may settle to Ethereum or to Celestia or to EigenDA, it keeps its own governance, and it owes nobody a royalty. That is a software relationship.
A chain that is part-of the Superchain is making a different claim. It is asserting membership in a federation with a shared bridge, a shared message-passing standard, a shared upgrade path, and a shared set of principles that constrain what it is allowed to do. That is a constitutional relationship, and constitutions have clauses.
Both models are legitimate. They are not the same business, and the industry's habit of comparing them on transaction speed is like comparing a franchise agreement to a software license by measuring the quality of the napkins.
What a stack actually inherits
In 2017, while I was supposed to be fixing bugs at a Swiss fintech, I spent three months reverse-engineering the Solidity contracts behind the Zeppelin security library and submitted four patches. The habit that survived from that detour is unglamorous: I read inheritance graphs. When someone tells me a chain inherits security, I want to see which contract it inherits from and who holds the admin key at the top of the tree.
So when I read the OP Stack's standard contract set, I look at four things.
The first is the standard bridge. It is genuinely shared code, and sharing it is a real safety gain: a canonical route between member chains means users are not forced through third-party bridges with their own multisigs and their own wrapped-asset accounting. That is the strongest technical argument for federation, and it is not marketing.
The second is the dispute game. Optimistic rollups resolve disagreements through a challenge process, and the interesting variable is whether that process is permissioned or permissionless โ that is, whether anyone can submit a challenge or only an allowlisted set of parties. A permissioned challenger set is not a bug in the design; it is an honest statement about what stage of decentralization the chain is at.
The third is the configuration contract. This is where the addresses live that can change the other contracts. It is the least glamorous file in the repository and the most important one.
The fourth is the upgrade authority pattern, and this is the piece I think is systematically underpriced.
Most federated rollups concentrate upgrade power in a security council with a delay window โ a group of named individuals who can push a change, subject to a timelock during which anyone can exit. This design is sensible. It is also, functionally, a mutualized risk pool. When you join the federation, you do not just inherit the code. You inherit the council's operational competence, the council's incident response culture, and the council's collective blast radius. If the council is asked to act fast during an exploit on one member chain, the precedent it sets touches every member chain.
That is the actual product being sold. Not throughput. Shared operational gravity, with a shared failure domain attached.
The charter is the product
In 2024, Optimism's governance adopted a document called the Law of Chains โ a set of principles that define what it means to be part of the Superchain. It covers things like maintaining Ethereum as the settlement and data availability layer, upholding user protections, and contributing to the collective.
Read it as engineering documentation and it is vague. Read it as a franchise agreement and it is precise.
A franchise is four things: a shared brand, shared standards, a shared supply chain, and a royalty. The Superchain has all four. The brand is the name. The standards are the bridge and the messaging format. The supply chain is the shared upgrade path. And the royalty is the revenue-share formula that member chains contribute to the collective out of sequencer revenue.
The royalty is the tell. It is also the part almost nobody discusses, because it is boring and because it does not fit into a thread about proofs per second.
Compare the alternatives on exactly this axis. Arbitrum Orbit charges no royalty and imposes no data availability requirement, which makes it the sovereignty-maximalist option. The ZK Stack sells shared interoperability through a gateway contract and governs through a token assembly. Polygon's AggLayer sells a unified liquidity and proof layer. Each of these is a different answer to the same question: how much of your chain's independence are you willing to trade for a shorter path to liquidity?
Notice what is not on that axis: ZK versus optimistic. The cryptographic debate is real, it matters for withdrawal times and for proving costs, and it will be settled by engineering. But it is not the axis on which the market is consolidating. The market is consolidating on the axis of federation membership, and the federation with the most compelling shared bridge will win on gravity rather than on merit.
Interop is the lock-in
The next phase of this war is already visible in every roadmap, and it is cross-chain messaging inside a family.
The mechanism is straightforward and the consequences are not. If two chains share a canonical bridge and a shared message bus, a user can move an asset between them without a third-party bridge โ no wrapped representations, no separate multisig, no separate trust assumption. The user experience improves and the safety surface shrinks. That is a genuine public good.
It is also a moat, and a durable one. Leaving the family means giving up the native route and reverting to the wrapped-asset, third-party-bridge world that everyone has spent five years learning to distrust. Switching costs in infrastructure are rarely technical. They are the accumulated weight of every integration that assumed the old route.
I spent a chunk of the 2022 bear market in Celestia's Discord servers arguing about data availability sampling and sharding economics with people who were building it. The lesson I carried out of those weekends was not about sampling. It was that shared security is always, without exception, a shared blast radius. A message bus that connects ten chains is a single surface through which a fault in one can propagate to ten.
Nobody prices this. Not because they are stupid, but because the failure mode is invisible until it is a headline, and because the people best positioned to price it are the ones selling the membership.
There is a second casualty worth naming. The independent shared-sequencer market โ Espresso, Astria, and the rest โ was built on the premise that ordering would be a neutral, cross-family service. If each federation ships its own interop layer with its own sequencer arrangement, that neutral market gets squeezed between the stacks. That is not a prediction of failure; it is a description of the box the market is now in.
The counter-intuitive part
Here is the claim that will annoy the most people: the stack that wins may be the one that convinces its members to surrender the most sovereignty, and that is not obviously good for the people who live on those chains.
The reasoning is not cynical. Users do not choose chains. They choose applications. Applications choose chains based on incentive programs and the cheapest path to liquidity. Liquidity follows the canonical bridge. So the federation with the largest shared bridge accumulates gravity, and gravity compounds without anyone making a good decision.
Which brings me to the chains that launch, distribute a token, run an incentive program, and then quietly stop mattering eighteen months later. The standard narrative is that they were abandoned, or that the team drained the treasury. Another rug pull? Or just another myth? Usually it is neither. Usually the mechanism is duller: the royalty was too high relative to the fee base, the bridge route was too indirect to attract real liquidity, and the incentive program was renting users rather than buying them.
And then there is the structural observation that I keep making and that keeps being received as an accusation. A security council with veto power over upgrades across a dozen chains is, functionally, a very small group of people holding a very large switch. That is true of every federated design, including the ones I think are well-run. It is not a criticism of any team. It is an observation about where the power actually sits once the marketing is stripped away.
The Cassandra complex is real. In 2020 I published threads about the impermanent loss traps in early liquidity pools and was told I was being negative about a sector that was obviously working. The traps were real. The people pointing at them were just early.
What to watch instead of throughput
Stop counting transactions per second. Start counting two numbers: how many chains share a single upgrade authority, and how long the delay window is on that authority.
Those two numbers tell you more about the risk you are actually carrying than any benchmark. They tell you how correlated your exposure is with strangers, and how much time you would have to leave if the council decided something you disagree with.
The question I cannot answer yet is the one that will define the next two years: when you sign a transaction on a chain you have never heard of, will you know whose council can pause it โ and will anyone have bothered to tell you before you clicked?