L1-zkEVM breakout #09, October 14, 2026

Agenda

Meeting Time: Wednesday, October 14, 2026 at 15:00 UTC (60 minutes)

GitHub Issue

YouTube recording available: https://youtu.be/VJnUl6udMGg

Meeting Summary:

The L1 zkVM breakup call focused on updates and progress related to zkEVM, execution layer (EL) clients, and proving infrastructure. Ignacio highlighted the upcoming ACDC call, where a decision on including CFI EIP-1525 into Hayabusa was expected, and shared updates on EIP-8025, EIP-7709, and recent test suite improvements. Cody’s video update covered expanded RISC-V compliance testing, a proposed standard for randomness in guest programs, and progress on ZISC verification. Francesco discussed consensus-spec refactoring, a draft EIP for recursive proofs, and plans to deploy approver on Sepolia. Manu reported on Prism’s proof-of-concept implementation of EIP-1525 and initial testing on Sepolia. The Silkworm team presented updates on their guest client’s performance gains, integration into the benchmark suite, and details about their optimized witness format (MFBD). Lion from Lighthouse shared that they had merged support for a proof engine sidecar and could run without an EL. The call concluded with a reminder about the importance of the upcoming ACDC decision.

Click to expand detailed summary

The team discussed updates on the L1 zkVM project, including highlights from recent developments and upcoming decisions. A key focus was the potential inclusion of EIP-8025 (CFI) into Hayabusa, with Ignacio encouraging strong opinions and support for the proposal in the upcoming ACDC call. The team reviewed progress on specs and tests, including recent refactoring work, the addition of new tests, and the creation of an R2 bucket for stateless inputs on Sepolia. They also discussed updates to the execution API, including the implementation of REST and SEC variants of the NewPayloadWithWitness API, and the creation of a new EIP for block building with witness.

Cody presented updates on RISC-V compliance testing, including additional coverage for accelerators and memory standards using the ACT4 framework, and proposed a standard to address a hash collision vulnerability in guest programs. He also discussed progress on ZISC compliance verification and updates to SoundCalc, including new STIR support. Ignacio then provided updates on the brewing infrastructure, highlighting Hans’ work on extending Benchmark Core to support proof generation and improve cluster management, with plans to implement a dashboard showing proving time, costs, and resource usage.

Francesco presented two main developments in the consensus specs, including a refactor of the proof data type and gosamp validation, and a reduction in the scope of the proof engine to focus exclusively on verification. He also introduced an initial draft EIP proposing an extension to allow a single proof to validate a chain subset, which would significantly reduce bandwidth requirements. The team plans to deploy on Sepolia to test the 8025 implementation, with an invitation for client teams to join the test network.

Manu reported implementing an EIP as a proof of concept on Kurtosis using Prism verification nodes, but encountered an incompatibility issue with Lighthouse peers due to proof type 0 being undefined. Francesco outlined two key open questions: the need to design a Beacon API format for Dora explorer integration, including methods for both submitting and fetching execution proofs, and discussions around container types for SSc in the consensus specs. Francesco is working on integrating Prism, Lighthouse, and their fork on a local devnet to resolve incompatibilities before deploying to Sepolia, and suggested rotating through different guest programs and provers for better coverage.

The team discussed updates on testing and proving SepoliaBlocks, with Ignacio agreeing to test as much as possible before going live. Francesco shared a config and slides link for review. Somnath introduced updates on Silkworm’s work, with Tullio providing details on Zilcor’s integration in the ERE benchmark. Tullio reported that their guest passed most tests, matched every Sepolia block, and was about half the cost of Ref’s proving cost. Future plans include native support for the canonical stateless input format in the next release (Alpha 5) and fixing the two failing tests.

Paweł presented performance gains achieved with Zilkworm Chrome between Alpha 2 and Alpha 4, including a 33% reduction in executed cycles and improvements in memory handling and byte swapping. Somnath discussed the development of z6mnode, a zkEVM execution client that achieved 2.2 seconds block timestamp to witness generation, and introduced the MFBD (Minimally Perfect Hash-Based Flat Witness Format) for optimizing witness data handling. Lion from Lighthouse announced that proof engine sidecar support was merged into Unstable, allowing Lighthouse to run with or without EL.

Next Steps:

  • ignacio: Share the links to Cody’s slides and his own slides in the chat.
  • Francesco: Share the slides link in the chat.
  • Francesco: Integrate PRISM and Lighthouse on a local devnet to iron out incompatibilities before deploying to Sepolia.
  • Francesco: Collaborate on deciding the Beacon API format for fetching execution proofs, and discuss container types for SSC.
  • Francesco: Organize a discussion on Discord about rotating the proof stack and guest programs for Sepolia testing.
  • Manu: Investigate the incompatibility issue with Lighthouse proofs of type 0.
  • ignacio: Share information on Discord about whether the next call will be skipped due to DevCon overlap.
  • All interested parties: Join the ACDC call tomorrow to support the proposal for CFI EIP-IT25 into Hayabusa.

Recording Access: