5 Steps to Make Ethereum Driven by LLMs
1/ Validator operators delegate to agents the decision-making on accepting/rejecting network upgrades and setting the parameters.
2/ EIP authors use LLMs to create and submit EIPs.
3/ EIP editors use LLMs to review and approve EIPs.
4/ All Core Devs use LLMs to moderate meetings and vote on EIP inclusion and spec changes.
5/ Client teams generate codebases from specs.
It is important for Ethereum to be the first chain to be LLM-driven; it is an advantage akin to being the first PoW chain. Ethereum has a natural advantage because it already has an existing spec that LLMs were trained on, and a very transparent governance process that LLMs can be trained on (all past ACD calls, EIP processes, open discussions) and participate in.
The EF recently hired tooling coordinators and established a dAI team. Between ACD moderators, tooling coordinators, Ethereum Cat Herders, EIP editors, and the dAI team, there should be a high-priority process on:
1/ Ensuring that agentic participation in the EIP submission process is simple and functional.
2/ EIP editors have tooling for AI review of all EIPs.
3/ There is AI support for real-time ACD moderation (connected to chat, analyzing the discussion content in real time, and making suggestions).
4/ Expanding forkcast.org over time to be a real-time, listener-context-aware broadcast of the Ethereum governance process.
5/ Establish a cross-client team of core devs that would work on an AI-generated client codebase driven by specs only. Such a client should be fully formally verified, test-covered, and developed in parallel to other codebases until it becomes a canonical client codebase.
1/ Validator operators delegate to agents the decision-making on accepting/rejecting network upgrades and setting the parameters.
2/ EIP authors use LLMs to create and submit EIPs.
3/ EIP editors use LLMs to review and approve EIPs.
4/ All Core Devs use LLMs to moderate meetings and vote on EIP inclusion and spec changes.
5/ Client teams generate codebases from specs.
It is important for Ethereum to be the first chain to be LLM-driven; it is an advantage akin to being the first PoW chain. Ethereum has a natural advantage because it already has an existing spec that LLMs were trained on, and a very transparent governance process that LLMs can be trained on (all past ACD calls, EIP processes, open discussions) and participate in.
The EF recently hired tooling coordinators and established a dAI team. Between ACD moderators, tooling coordinators, Ethereum Cat Herders, EIP editors, and the dAI team, there should be a high-priority process on:
1/ Ensuring that agentic participation in the EIP submission process is simple and functional.
2/ EIP editors have tooling for AI review of all EIPs.
3/ There is AI support for real-time ACD moderation (connected to chat, analyzing the discussion content in real time, and making suggestions).
4/ Expanding forkcast.org over time to be a real-time, listener-context-aware broadcast of the Ethereum governance process.
5/ Establish a cross-client team of core devs that would work on an AI-generated client codebase driven by specs only. Such a client should be fully formally verified, test-covered, and developed in parallel to other codebases until it becomes a canonical client codebase.
Thanks to @trent_vanepps , @dapplion , @austingriffith , @ralexstokes , @hasufl , @ergokhaner and others for both the brief and longer conversations over the last days.
@trent_vanepps @dapplion @austingriffith @ralexstokes @hasufl @ergokhaner Also a longer convo about the future of Ethereum with @vinayvasanji , @rudolf6_ , @owocki , @dannyryan , @0xstark - inspiring.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
