Post-Quantum (PQ) Interop #44, June 17, 2026

Protocol Calls & happenings
1

Agenda

  • devnet5
  • leanspec refactor

Meeting Time: Wednesday, June 17, 2026 at 14:00 UTC (60 minutes)

GitHub Issue

2

Meeting Summary:

The meeting focused on client updates and progress on DevNet 5 implementation across different teams. Gajinder from the Zim team reported working on stability fixes and proof merging optimization, noting that proof aggregation was taking too long on standard machines and requiring optimization or hardware upgrades. Shariq from the Veeam team discussed memory usage improvements and plans to move to DevNet 5 after fixing resource-intensive issues in DevNet 4. Mihir from Lantern reported finding the root cause of snappy decoding errors and working on block building optimizations, achieving 0.9-second block proposal times. Pablo from Lambda mentioned merging unified support for membranes and implementing cross-plane metrics. Boma from Grandin reported timing issues with Type 2 final block proofs exceeding 4-second slots, leading to discussion about potential solutions including hardware upgrades, timing adjustments, or proof efficiency improvements. Thomas provided updates on LeanSpec simplification efforts, including audits, documentation improvements, and the addition of multilinear proving system skeleton to Plonky 3. The team also discussed a PR from Lambda that changes the block building algorithm to potentially address proof timing issues, with agreement that while it adds complexity, it could be valuable for clients to explore.

Click to expand detailed summary

The team provided updates on client development and DevNet 5 implementation. Gajinder reported progress on proof merging optimization, noting that on a Core i7 machine, proof merging now takes 0.4-0.5 seconds, and discussed plans for type one aggregation and split proof mechanisms. Shariq shared that the Veeam team is working on stability fixes and resource optimization for DevNet 4, addressing memory usage issues up to 28 GBs, before transitioning to DevNet 5. Mihir reported fixing the snappy decoding error in Xeem and implementing support for aggregator skipped metrics, while working on block building optimizations for DevNet 5.

Mihir explained that the 0.9-second block proposal metric in DevNet 5 measures only block building time, not including the final block proof, and he is currently focusing on optimizations for block building rather than aggregator duty. Pablo reported that the team merged unified support for membranes, aligned on choice finality and aggregate validation, implemented cross-plane lean aggregator metrics, and began working on a Goldfish prototype for potential implementation in DevNet 6 or a future event.

Gin reported that DevNet 5 focused on updating Gene to align with the latest lean spec changes, addressing areas where behavior could diverge from the specification to ensure consistency with other implementations and reduce interoperability risks. The team plans to continue testing against live multi-client interop while tracking specification updates. Boma discussed issues with QFix gossip and Type 2 final block proof stack timing, noting that current builds exceed the 4-second slot budget, which is affecting proposal finalization.

Boma and Gajinder discussed issues with Type 2 proofs taking longer than expected during DWNet5 testing, currently taking around 2.6 seconds. Gajinder proposed either upgrading hardware to Core i7, improving Type 2 proof efficiency, or increasing the time slot to 12 seconds. Gajinder expressed optimism about potential optimizations that could help Type 2 proofs meet the 0.8-second requirement within the current hardware constraints.

The team discussed optimization options for proof generation, with Boma noting they reduced attestation pack entries from 8 to 4 while maintaining the same deadline. Thomas explained that while some improvements are possible, they should not expect a 2X speedup and suggested discussing options like increasing machine requirements, extending slot time, or improving proof efficiency with Emil in follow-up discussions. The team also reviewed recent progress on VIN specs simplification and Plonky 3 development, including the addition of a multilinear proving system skeleton, and identified proof timing as a key focus area for DevNet 5.

The team discussed a PR (791) submitted by Lambda that changes the block building algorithm to address performance issues with block proposal times. While the PR adds complexity to the spec, the group agreed it could be useful for client teams to explore, with the possibility of keeping it as a separate optimized function rather than merging it directly into the spec. Thomas and Gajinder emphasized the need to keep the spec simple while allowing for optimization strategies at the client level, including implementing strict timeouts and prioritization mechanisms for node work.

Next Steps:

  • Thomas: Prepare and share slides about LeanSpec simplification progress and future direction with the EF team by end of week or beginning of next week.
  • Thomas: Post a message on the channel for Emil to follow up on expected optimization margins for Type 2 proofs, and discuss options (machine requirements bump, slot time increase, or proof efficiency improvements).
  • Gajinder: Continue working on and testing the optimization PR for starting Type 1 aggregation in the previous slot interval, and debug/test the split proof mechanism for attestations on non-main branches.
  • Gajinder: Continue integrating Zig libp2p and debug interoperability issues to achieve stable interop for Zeem.
  • Shariq: Fix excessive memory usage, finalization delays, and justification delays in Veeam on DevNet 4 before migrating to DevNet 5.
  • Mihir: Test block building optimizations with a larger number of validators and nodes, and continue exploring pathways to speed up aggregation duties for Lantern on DevNet 5.
  • Pablo: Continue working on the Goldfish prototype proposal for DevNet 6 or a future event.
  • bankat: Continue testing Gin against the live multi-client interop on DevNet 5, track specification updates, and stay aligned with development.
  • Boma: Investigate and address Type 2 final block proof timing issues observed during DevNet 5 local runs.
  • All client teams: Review Lambda’s PR #791 on the block building algorithm change to explore capping attestation data and minimizing aggregation time during block building.
  • Mega: Keep PR #791 available for client teams to review, and consider restructuring it as a separate optimized block building reference rather than merging it directly into the main spec.

Recording Access:

3

YouTube recording available: https://youtu.be/TvxI5kJfx4Y