Ethereum Co-founder Vitalik Proposes Recursive STARK Memory Pool EIP-8288
Ethereum co-founder Vitalik Buterin introduced EIP-8288, aiming to incorporate a recursive STARK memory pool in the upcoming I-star upgrade after Hegota. This proposal seeks to move signatures and proofs out of the core execution path, aggregating them in the memory pool before being added to the blockchain.
Transactions carry dependency frames, declaring "certain signatures or conditions are met," while the original large proofs remain off-chain. Nodes periodically consolidate non-expired transactions, generate recursive STARKs, and broadcast the encapsulation; block builders then produce the final proof for the transactions intended for packaging. Each blockchain will bear an additional burden of about 100 to 300KB for STARKs, plus 96 bytes for each statement to be proven. The gas cost for quantum-safe privacy transactions is expected to drop from around 10 million to tens of thousands. The proposal allows for the integration of post-quantum signatures like Falcon and ML-DSA without modifying the EVM, and supports private account abstraction. The candidate instruction set for recursive proofs points to RISC-V. The proposal was co-drafted with Thomas Coratger and is built on EIP-8141 frame transactions, currently in draft form.
The bandwidth logic involves nodes exchanging transaction bodies along with an aggregated proof, rather than attaching a complete post-quantum signature to each transaction. Builders only need to verify the final block collection once, avoiding the expensive verification for each transaction at the execution layer.
In terms of market mechanisms, this shifts the proof cost from the execution layer to the memory pool and builders. Beneficiaries include privacy protocols, post-quantum wallets, and account abstraction wallets; those under pressure are the old precompiled paths that still charge per transaction for verification, as well as nodes and relays that must upgrade their provers. The decrease in gas costs will change the frequency of sendable privacy transactions but will not directly alter ETH issuance.
The scope of I-star has not yet been locked down. The draft can be included in a fork, depending on whether clients and builders accept the engineering burden of RISC-V proof loops.
Source: Public Information
ABAB AI Insight
The large size of post-quantum signatures makes directly integrating them into the EVM a luxury for privacy and quantum resistance. Vitalik chose to modify the memory pool instead of the virtual machine: shifting verification from "executed once by each node" to "the network first recursively compresses, leaving only one proof on-chain." This is the next step after Frames, allowing the execution layer to continue running old contracts while cryptography evolves through a sidestep. RISC-V is named as the standard machine for proof statements, effectively reserving the same circuit for all future new algorithms instead of adding a new precompiled for each new algorithm.
The capital implication lies in L2 and wallets: whoever first adopts the dependency frame can turn the million Gas privacy path into a regular transaction. Builders gain access to a new proof power market, and relays must be able to verify STARKs to forward them. Coinbase's Stanford quantum-resistant discussions focus on Bitcoin output types, while Ethereum is pursuing memory pool recursion, with both chains addressing the same physical threat through different interfaces.
In contrast to EIP-4844, which dumps data into blobs, and various signature aggregations, the industry is shifting from execution optimization to proof optimization: the expensive part is no longer the computation itself, but who has the authority to fold the computation into a block header attachment.
This represents a technical substitution: on-chain per-transaction verification is replaced by memory pool recursive proofs. The mechanism is that the volume issue does not enter the EVM, the timeline is tied to I-star, the instruction set is tied to RISC-V, and privacy and quantum resistance share the same pipeline.
ABAB News · Law of Cognition
When signatures are too expensive, first modify the memory pool, not the virtual machine.
A single proof can cover the verification of an entire block, allowing gas costs to drop from millions to tens of thousands.
New algorithms can be integrated because the proving machine is first defined as the same set of instructions.