Meeting Summary:
The AllCoreDevs meeting focused on the execution of hard fork 247 on Sepolia testnet and planning for the upcoming hoodie fork. The team successfully completed the Amsterdam fork on Sepolia with gas limit increased to 200 million, though builder participation remained low with only 4 active builders. For the hoodie fork scheduled for October 26th at 1742 UTC, there was discussion about the tight timeline for client releases by October 16th, with concerns raised by POTUS about the short notice for large staking pool operators. The team reviewed 11 EIPs requiring decisions, with several being moved to DFI (deferred for future consideration) including EIPs around mempool improvements and precompile changes. Dan presented a complexity assessment tool using LLMs to evaluate EIPs, showing correlation between accumulated complexity scores and time from DevNet to mainnet deployment. The meeting also covered discussions about accelerating development processes and improving parallelization of work across multiple EIPs, with emphasis on using AI tools for initial assessments while maintaining human oversight of critical decisions.
Click to expand detailed summary
The team discussed the successful completion of the Amsterdam fork on the Sepolia testnet, with gas limits now ramped to 200 million. Barnabas reported that only 4-5 builders are currently active on Sepolia, primarily from Titan and Sigma Prime, with limited participation from major players. The team then moved to discuss scheduling the hoodie fork, with nixo presenting two potential time slots for the last week of October (Monday the 26th and Tuesday the 27th), though the final decision was not made during the meeting.
The team discussed scheduling a fork for DEF CON, with initial proposals for Monday the 26th at 11 UTC and later consideration of Tuesday at 8 PM UTC. After debate about timing constraints, including travel schedules and release preparation needs, the group decided to schedule the fork for Monday the 26th at 11 UTC. The team also confirmed that gas limit changes would be implemented on the fork epoch for Hoodie and Mainnet.
Stefan reported on the mainland Shadow Fork implementation, which successfully tested fork transitions and large block processing, though the benchmarking aspect had issues with gas limits. He plans to relaunch the benchmarking Shadow Fork next week with improvements. Dan presented a retrospective study on EIP complexity assessments using LLMs, showing a correlation between accumulated complexity scores and time from DevNet to mainnet across previous forks. The team discussed using this tooling to help EIP authors improve specifications rather than focusing on the complexity numbers themselves.
The team discussed implementing and assessing EIPs (Ethereum Improvement Proposals) using AI tools like LLMs, focusing on balancing automation with human oversight. Dan suggested creating a dashboard to validate specs and assess their implementation complexity, while Łukasz highlighted the need for better tools to streamline coordination, testing, and involvement of EIP champions. The group agreed on the importance of maintaining human review and steering while leveraging AI for efficiency, and Justin emphasized the need for specific goals when accelerating EIP implementation. Open questions included how to ensure spec quality and client diversity in interpretations, and how to balance automation with human involvement in the process.
The team discussed client diversity in specifications, with Łukasz clarifying that client diversity should not be considered in core specifications as clients need to agree on core specs. The group reviewed multiple EIPs for the upcoming Hegota fork, making decisions on which ones to CFI (Community Finalized Initiative) versus DFI (Developer Finalized Initiative). Key decisions included CFIing several EIPs including 5920, 8151, 8298, and 7709, while deferring others like 7666 and ML-DSA verification due to lack of consensus. The team also discussed the potential addition of a new EIP regarding contract size metering, which Ben Adams proposed for reconsideration.
Next Steps:
- Stefan: Relaunch the benchmarking Mainnet Shadow Fork at the beginning of next week, ensuring it runs more smoothly.
- potuz: Add a flag to Prism to disable the P2P builder stack and allow fallback to trusted builders/local building, ready for the Hoodie fork.
- All clients: Release client versions for the Hoodie fork by October 16th.
- nixo: Confirm the gas limit change will occur at the fork epoch for both Hoodie and Mainnet.
- danceratopz: Continue to iterate on the complexity assessment tool, focusing on providing signals for spec validation and contradictions.
- Łukasz: Prototype tools to streamline EIP implementation, testing, and feedback loops, potentially involving LLMs for automated PRs and dashboards.
- Ben Adams: Reconsider and potentially CFI the EIP for metering based on contract size (related to increasing contract size limits) for the next scoping call.
- nixo: Add the new EIP (metering based on contract size) to the PFI list for consideration in the next call.
- nixo: Prioritize the CFI EIPs for Hegota, with 5920, 8151, 8298, 8360, 7709, and 8116 included, and 7666 DFI’d.
- Pooja: Add the Glamsterdam Ecosystem Readiness Checklist for Devs Attention to the Ethereum PM repo and share it with enterprises for feedback.
Recording Access:
