Post-Quantum (PQ) Interop #46, July 1, 2026

Protocol Calls & happenings
1

Agenda

  • devnet5
  • binary fields/Flock

Meeting Time: Wednesday, July 01, 2026 at 14:00 UTC (60 minutes)

GitHub Issue

1 Like
2

Meeting Summary:

This meeting focused on client updates and progress on DevNet 5 testing, with participants sharing their recent developments and ongoing work on performance optimizations. Gajinder from Zim reported that their client is now stable for multi-devnet and multi-subnet networks, having successfully run interop tests with 4 clients and 8 subnets, and proposed a two-level aggregation system to improve performance by increasing intervals from 5 to 8 and introducing type 1 and type 2 aggregations. Rim from Reem shared updates on PQ interop testing with 24 validators achieving up to 90,000 slots but experiencing some delays in finalization. Other participants including Ruslan from Clean, Mihir from Lantern, Pablo from Islambda, and Bankat from Jean provided updates on their respective clients’ performance optimizations and spec catches. Thomas provided an update on cryptographic work, mentioning the publication of a new binary fields proving system called Flock that can process around 600 Blake3 hash functions per second, approximately 2x slower than Poseidon. The team discussed testing optimization proposals from East Lambda and Zim, with Mega clarifying that the first two optimizations should remain client-side while the third requiring spec changes needs further consideration. Parthasarathy reported running a 32-node public DevNet with 4 clients showing stable justification and finalization, though some issues with Reem’s image compatibility were identified and being addressed.

Click to expand detailed summary

Gajinder reported that Zoom has achieved stability and robustness for multi-devnet and multi-subnet network participation, successfully running interop tests with four clients and eight subnets. He explained that delays in attestation state and block proposal are causing intermittent finalization issues, and proposed a two-level aggregation system to address these problems. The proposal includes removing type one recursive aggregation from the proposal pipeline and adding three intervals to the block proposal process, with details available in a GitHub issue.

Gajinder proposed increasing the aggregation intervals from 5 to 8 intervals, which would result in 0.4-second slots for type 2 aggregation. While the team has not tested this idea yet, Gajinder explained that type 2 aggregation currently causes slowdowns, so they are proposing to align type 1 recursive aggregation with safe target calculation intervals and remove the unique attestation data constraint. Gajinder also mentioned that shadow Sims have gained stability after recent fixes, and they plan to run a 32-node or 16-node network for analysis using Camille’s visualization tool.

Rim provided an update on PQ interop testing with 24 validators on DevNet, achieving 90,000 slots but experiencing a 30-slot delay in finalization due to attestation time. Ruslan reported limited progress on Clean, with some high tests still not passing. Mihir announced the addition of an ETH Lambda-style optimization for block building and support for a simplified block building PR, along with a new shadow simulator image compatible with Camille’s lean shadow fuzzer tool.

Pablo reported changes to history pruning for minimal disk usage and ported ETH P2P to Rust, while also sharing a document on cell duration bumping. Bankat updated on Jean’s progress, including fixes for attestation committee count and spec updates, along with improvements to the Shadow prover cost model. Boma discussed cleanup PRs and Shadow support, while Gajinder clarified that the proposed shared source recognition change in SDF involves vote propagation up the ancestor chain to facilitate faster justification without significant changes across client codebases.

Anshal proposed changes to the attestation process to address long delays in finalization and justification by considering implicit votes when source and target blocks lie on the same chain. Gajinder discussed computational concerns and block size limitations, clarifying that while multiple proofs per data are allowed, the current system still maintains bounds on block size. Thomas provided an update on current work, including testing new proposals and implementing binary field proving systems, with plans to stabilize the spec while focusing on performance improvements and bug fixes.

The team discussed client optimizations, with Mega explaining that three optimizations were shared in a document, with only one potentially requiring specification changes. Parthasarathy reported on DevNet testing results, noting that a 12-hour test with 32 nodes and 4 clients ran successfully with no justification issues, though finalization occasionally took up to 45 minutes due to different client aggregation methods. The team identified some issues including ARM compatibility problems with Reem’s image and interop issues between Reem’s Ziglip B2P and Lantern, which have been fixed. Thomas suggested continuing to test optimizations and debug DevNets while increasing validator numbers, with Kolby and Derek agreeing to investigate Hive’s failures which appear to be caused by using an old spec version.

Next Steps:

  • Gajinder: Test the two-level aggregation proposal in the coming week if more people are aligned.
  • Gajinder: Run a 16-node or 32-node shadow sim network and use Camille’s visualization tool to analyze and present findings in the next call.
  • Parthasarathy: Restart the public DevNet with all six clients after fixing the interop issue between Reem’s latest Ziglip B2P and Lantern.
  • Parthasarathy: Confirm the availability of the official REAM image with the team for inclusion in the DevNet.
  • kolby and Derek: Look into the Hive failure issue today, as it may be using an old version of the spec.
  • Thomas: Consider redesigning the PQ interrupt call to include more cryptography and possibly remove client updates in favor of presentations.
  • All clients: Test the optimization proposals (from ETH Lambda and Zim) in the coming weeks.
  • All clients: Continue to debug the DevNets and increase the number of validators as debugging progresses.

Recording Access:

1 Like
3

YouTube recording available: https://youtu.be/x5Z6gtkxnAc