Starknet achieves 2.1 GiB peak memory for mobile STARK proof
Summary
In a breakthrough collaborative research effort, the peak memory required for the STARK proof generation by the Stwo prover dramatically decreased from 13.9 GiB to just 2.1 GiB in under two weeks, showcasing the efficiency of combining human and AI input. This optimized prover allows mobile devices, like the Pixel 11, to generate a full STARK proof for a Starknet mainnet transaction in only 51 seconds, thereby enhancing privacy by enabling proof generation locally on user devices rather than on servers that can access transaction details. This hash-based STARK approach offers post-quantum security without the need for trusted setup, marking a significant advancement over traditional elliptic curve systems used in other privacy protocols.
Tokens
$ZEC$RAIL
Analysis
Starknet: Starknet is a decentralized layer-2 scaling solution on Ethereum that relies on STARK proofs for verifiable and scalable computation. The network's mainnet transactions were central to the recent optimization of the Stwo prover, enabling full proofs to run directly on mobile devices. Omar Espejel: Omar Espejel contributed to the public workshop on Stwo prover optimization and shared key details about the mobile proof generation milestone. He highlighted the collaborative process involving humans and AI agents on the project. StarkWare Ltd: StarkWare Ltd develops STARK-based proving systems and powers the Starknet ecosystem with its technology stack. The company's Stwo prover underwent rapid efficiency gains through open collaborative research highlighted in the news. Privacy: STARK proofs enable private transaction flows by allowing proof generation to occur locally on user devices rather than on servers that access transaction details. Security: Hash-based STARK proofs provide post-quantum security at the proof layer with no trusted setup required, in contrast to elliptic curve systems used in some other privacy protocols. Collaboration: Open research workshops bring together humans and AI agents to tackle complex problems like prover optimization through public threads, idea graphs, and shared experimental results.
Categories
ethereumai_agentsaidefitech