Ethereum co-founder Vitalik Buterin has shifted from strong skepticism to strong support for a technical approach that could accelerate the network’s path to quantum resistance while limiting the cost impact on ordinary users.
In a public post on November 16, 2026, Buterin wrote that over the preceding weeks he had moved from viewing the “leanPCS” method as unworkable to regarding it as highly promising.
He framed the change as opening a faster route toward what he described as a cryptographic proof singularity—an environment in which large-scale verification of signatures and other heavy computations becomes practical without permanently storing the underlying data on the chain.
The proposal at the center of the discussion is Ethereum Improvement Proposal 8288, titled “In-mempool signature and proof aggregation.”
It builds on earlier frame-transaction designs and aims to handle any computation that must be performed once, after which the chain only needs to record the result.
Signatures, zero-knowledge proofs, Merkle branches, and certain fully homomorphic encryption steps all fall into this category.
The two most consequential cases for near-term quantum safety are signatures and STARKs.
Current Ethereum account signatures based on ECDSA occupy roughly 65 bytes.
Quantum-resistant alternatives such as Winternitz one-time signatures or SPHINCS-family schemes expand that figure to between one and six kilobytes depending on parameter choices.
Zero-knowledge proofs face an even larger expansion: constructions used today can occupy a few hundred to a thousand bytes, while transparent STARK proofs can reach approximately 128 kilobytes.
Placing either category of object directly into every block would rapidly consume scarce block space and raise fees for ordinary transfers and application interactions.EIP-8288 addresses the problem by moving aggregation into the mempool.
A user still generates the full signature or proof and publishes it once.
Nodes then collect newly seen objects and, at regular intervals, produce a recursive STARK that attests to a compact list of claims—for example, that a given public key signed a particular message, or that a specified program produced a stated output on given data.
Only short hashes of the inputs and outputs (roughly 64–96 bytes per claim) together with a single aggregate proof are rebroadcast.
These aggregate proofs can themselves be recursively combined until a block contains one master proof covering every included claim.
The permanent on-chain footprint is therefore limited to the claims themselves, stored in a specialized frame type, rather than the original bulky objects. Buterin contrasted three concrete ways the framework could be realized.
The first, leanSPHINCS, relies on a purely hash-based signature scheme.
It would let users create quantum-safe accounts and multisignature arrangements while paying only a modest gas cost—on the order of a few thousand gas for the short claim data.
Its limitation is its narrow scope: it supports only that particular signature algorithm.
The second option, leanSTARK, treats the user-generated object as a general STARK.
Because any computation can be proved, the approach supports arbitrary signature schemes (including lattice-based ones), privacy protocols, and private account-abstraction logic in which authorization depends on hidden state.
The practical obstacle is that the protocol would need to standardize an instruction-set architecture for the programs users write.
Agreeing on such a language, which would be difficult to upgrade later, is expected to take considerable time.
The third route, leanPCS, rests on polynomial commitment schemes—the core hard component of STARKs.
A PCS is simply a proof that a polynomial evaluates to a claimed value at a given point.
From the perspective of the Ethereum Virtual Machine, EIP-8288 supplies what amounts to a “magic” PCS oracle; once that oracle exists, verifying a full STARK can be reduced to roughly 20,000–40,000 gas.
Specifying a PCS requires far fewer design choices than a complete instruction set; the main decision is the choice of finite field, for which a natural candidate already exists (a 128-bit binary field, or a recursive tower of binary fields up to 256 bits).
On-chain, the method is somewhat less efficient than a pure leanSTARK because multiple claims may be needed for a single logical proof, yet it offers greater flexibility in the off-chain proof systems that can be used.
Most importantly, it delivers essentially the same generality as leanSTARK while appearing implementable on a shorter timeline.
Once an instruction-set architecture is later agreed upon, a full leanSTARK mode can be added.
Buterin concluded that the combination of EIP-8288 with leanPCS therefore supplies a practical path to the desired cryptographic capabilities that is faster than previously expected. The proposal remains in draft status, with a reference implementation still incomplete and no firm activation schedule attached.
