Post Quantum transaction signature (PQTS) Breakout #16

Protocol Calls & happenings
1

Meeting Time: Wednesday, September 30, 2026 at 13:00 UTC (60 minutes)

GitHub Issue

1 Like
2

YouTube recording available: https://youtu.be/hXG-ahMY2oA

3

Meeting Summary:

This meeting focused on discussing the implementation of EIP-8288 for post-quantum signature aggregation in the Daisugi testnet. Antonio led the discussion about moving from testnet version 02 to version 03, which involves implementing signature aggregation using LinVM and LinSphinx. The team explored how to handle frame transactions and verification processes, with Matteo Vicari explaining that they are using Flock and Wir for wrapping proofs in LinVM. Stefano De from Nethermind provided insights on the verification process, clarifying that native verification inside Nethermind would be needed without relying on precompiles or smart contracts. The group discussed performance considerations, with current aggregation rates around 20 signatures per second, and explored how to handle the verification of aggregated signatures in the mempool. The team agreed to continue collaborating on implementing the aggregation mechanism and verification processes in the next version of the testnet.

Click to expand detailed summary

Antonio announced the launch of a testnet called Daisugi, which includes two versions: version 0.1 with non-native account abstraction and version 0.2 with options for ERC4337 and EIP8141. He mentioned that the next phase involves implementing signature aggregation in the mempool using LeanVM and LeanSphinx, and sought help from participants, including Stefano and Tomana, for this effort.

Antonio discussed the development of testnet versions, highlighting the progression from version 01 with non-native account abstraction to version 02, which includes both non-native and frame transaction options. He explained the plans for version 03, which involves moving into “virgin territory” with aggregation features using LIN Sphinx and LIN VM. Antonio sought additional help for this effort and attempted to use a whiteboard to facilitate the discussion, though technical difficulties prevented its display.

Antonio and the team discussed challenges with using whiteboards in Zoom and explored alternative solutions, with Matteo suggesting Excalidraw as a viable option. The main focus of the meeting was Antonio presenting a new IP proposal for a block construction and aggregated signature system, which involves introducing an aggregator role to manage transactions and signatures more efficiently, particularly in the post-quantum case. The proposal includes modifications to the NetherMind client to incorporate LinVM for verifying aggregated signatures, with Antonio explaining that this approach would maintain a constant size for aggregated signatures regardless of the number of transactions.

The team discussed the implementation of aggregation using Flock and Weir for wrapping proofs in the LeanVM repository. Matteo Vicari explained that the current proof size is approximately 200K and processing time is around 20 signatures per second on a regular machine, though these numbers are preliminary and expected to improve. Tomás shared that they created a prototype for in-mempool aggregation with LeanVM in ethrex, focusing initially on validation rather than more complex features like recursive aggregation, which will be implemented in future iterations.

Antonio and Matteo discussed the implementation of EIP-8288, focusing on type 7 test cases involving a wrapper with deep VerifyFrame mode. They explored how frames would work, including triples containing message hashes and public key hashes, but acknowledged the complexity of the implementation and the need for further clarification from Giulio and potentially Vitalik. The team recognized that while they had started working on the account abstraction part, which was easier to understand, the lean implementation would require more time and effort to fully comprehend.

Stefano asked about the typical number of signatures aggregated in a block for Lean Sphinx, and Antonio clarified that it ranges between 200 to 2,000, typically around hundreds. The team discussed performance metrics, with Antonio noting that the current version of Lean Sphinx achieves around 180 verifications per second, significantly lower than the original version’s 280 per second. The discussion highlighted that performance varies based on parameters like the hash function used for aggregation and the specific implementation details.

The team discussed implementing frame transactions with LinVM for signature aggregation, focusing on test case 7 which involves verifying LinVM proofs natively within Nethermind without requiring precompiles or smart contracts. Stefano De explained the high-level architecture where frame verification checks the dependency structure while the wrapper object contains the actual signature data, but the team identified challenges in implementing the verification process efficiently within the client. Antonio invited Stefano to join the testnet development effort, specifically for version 0.3 of the testnet using Nethermind, where they are working on implementing the missing aggregation functionality for EIP-8288.

Next Steps:

  • Antonio: Add Stefano De to the Daisugi testnet Telegram group for collaboration on version 0.3.
  • Stefano De: Work on a concrete solution for EIP-8288 test case 7 and bring it to the next call.
  • Antonio: Try to bring Vitalik or Thomas (authors of EIP-8288) to a future call for clarification.
  • Antonio: Continue to seek additional help for the testnet version 0.3 implementation of EIP-8288.

Recording Access: