Vitalik Buterin said on Monday that he now favours a design called leanPCS for EIP-8288, a draft improvement proposal he co-wrote with Thomas Coratger in June to keep quantum-safe signatures from swamping Ethereum’s blocks. In a thread on X, he compared three routes and argued the polynomial-commitment version could ship sooner because it only requires picking a mathematical field.
The size problem is the reason the proposal exists. Ethereum accounts sign with ECDSA today, at 65 bytes per signature. Quantum-safe hash-based signatures would take 1 to 6 kilobytes, and zero-knowledge proofs written as STARKs could grow to 128 kilobytes. Buterin’s assessment of putting that on chain was blunt: it is unsustainable, and 8288 is meant to fix it.
The fix moves verification off the chain’s storage and into the mempool, where pending transactions wait. Nodes check users’ signatures and proofs there and fold them into recursive STARKs, so each block carries a single aggregate proof rather than every signature. Only the statements those proofs vouch for — which public key signed which message — reach the chain.
The three options differ in what they ask the protocol to accept. leanSPHINCS is a single hash-based signature scheme, built on the SPHINCS+ family that NIST standardised in 2024, and it would let users hold quantum-safe accounts for very little gas per signature. Its limit is coverage: it handles one specific scheme. leanSTARK would accept a proof of any computation, but only once Ethereum settles on an instruction set architecture, which Buterin called a really big deal and likely to take time to agree on. leanPCS, the option he now prefers, lets Ethereum verify a STARK for roughly 20,000 to 40,000 gas.
His change of mind was not gradual. “Over the past few weeks I’ve gone from thinking leanPCS is total bull to being a total bull,” he wrote, describing it as a path to what he called the cryptographic proof singularity arriving much faster than anyone expected.
None of this is scheduled. EIP-8288 is a draft, and activating it would require a network upgrade with everything that entails. But the sequencing argument is the substantive part: a scheme that needs only a field choice can be specified and tested now, while leanSTARK waits on an ISA decision that has no date. Layer 2s are already moving on the same question, with Starknet weighing a move to layer 1 to pursue quantum resistance by 2027.