The term “Ethereum layer-2” is generally thought to mean that a chain inherits Ethereum’s security. But the line has always been blurry, and remains so, especially with the rise of OP Stack chains that use alternative data availability (DA) layers. These so-called “Optimiums” settle to Ethereum but post data elsewhere, introducing some additional trust-assumptions.
Toghrul Maharramov, a rollup researcher and former Scroll contributor, has been a vocal critic of looser L2 definitions, and declared Wednesday on X:
“Optimiums are not L2s. Anyone who says otherwise is trying to psyop you.”
A good example of this is Celo, which Blockworks reported as a case study of an L1 becoming an L2. For the most part, Ethereum rollups post their data to Ethereum itself — originally via calldata and now, predominantly, via blobs. That enables the useful property of permissionless state reconstruction and withdrawals from an L2 to Ethereum mainnet.
Optimiums, by contrast, outsource DA to off-chain networks like EigenDA, Celestia or Avail, and rely on attestations to prove that data is available. They may use fraud proofs and settle to Ethereum, but they add a layer of trust that Ethereum cannot directly verify.
Vitalik Buterin has used “L2” liberally, and indeed specifically referred to Celo as moving to an “L2,” so we’re in good company there, but the difference is more than semantics.
Ethereum scaling tracker L2BEAT now classifies most Optimiums as “Others,” a catch-all category for chains that don’t meet the strict rollup criteria. Celo, for example, completed its transition in March 2025 to run on the OP Stack with EigenDA.
According to L2BEAT, it fails two key checks for rollup status:
- It has fewer than five external actors who can submit fraud proofs
- It has no DA bridge that posts EigenDA availability proofs on Ethereum
The result: no L1-enforced remedy if data goes missing.
Lisk, which migrated to Ethereum in late 2024, also falls into this bucket for similar reasons — reliance on a permissioned challenger set — although it does post data on Ethereum.
According to Sreeram Kannan, founder of EigenLayer, we can think of rollups as “out-of-protocol scaling for execution,” while EigenDA is “out-of-protocol scaling for blob space.” In both cases, the data and logic happen outside Ethereum’s protocol rules — but only rollups ensure Ethereum can enforce validity directly.
“When a certificate from EigenDA shows up, you’re trusting that that certificate implies that the data is available,” he told Blockworks.
That trust model has no on-chain enforcement today.
“If the DA claimed the data is available and the data is not available, there’s no recourse,” Kannan said.
So, is “L2” just Ethereum-aligned marketing? Or should it reflect the protocol’s ability to enforce both correctness and availability?
L2BEAT’s current stance is that Ethereum must be able to reject bad state roots and verify data availability independently. Anything else requires more trust — whether that’s in a DA quorum, a governance council, or a permissioned set of challengers.
Optimiums without either state validation or a DA bridge remain in the “Others” category. But as the tech matures, some may graduate to “Validiums & Optimiums” or even gain rollup status. To do that, you’d need some robust fallback mechanism or a way to make DA failures an “objectively attributable fault,” — a tough problem to crack.
Until then, the debate continues, grounded in practical considerations for users about who is trustworthy to keep funds safe.
“If you’re just honest, every layer-2 is controlled by a multisig, so we should be talking more about that,” Kannan said. This is certainly a prime bugaboo for critics of Ethereum’s scaling scheme, he acknowledges.
“But I would want to also point out that every app on Solana has like a 2-of-3 or some very small multisig that controls it,” he said, meaning there’s more work to be done to minimize trust all around.
Get the news in your inbox. Explore Blockworks newsletters:
Source: https://blockworks.co/news/optimiums-and-the-l2-ethereum-security-revisited