Tomasz Stańczak, co-director at the Ethereum Foundation, has published a detailed five-step blueprint to transform Ethereum into the first blockchain governed by large language models (LLMs). The plan, outlined in a post on X, aims to leverage AI to manage network upgrades, improve proposal processes, and even generate client code, positioning Ethereum with a strategic advantage in the emerging AI-powered blockchain race.
The first step involves validator operators delegating decision-making authority to AI agents for tasks like approving network upgrades and setting parameters. Stańczak argues that being first to adopt AI governance could provide Ethereum with a competitive edge similar to its early lead with proof-of-work. The subsequent steps are structured to integrate AI across the entire governance stack: EIP authors would use LLMs to draft and submit proposals; EIP editors would employ AI review tools; All Core Developers would utilize LLMs to moderate meetings and vote on EIP inclusions; and finally, client teams would generate entire, formally verified codebases directly from specifications.
The proposal is not merely theoretical. The Ethereum Foundation has already hired tooling coordinators and formed a dedicated dAI (decentralized AI) team to drive implementation. Stańczak emphasized that Ethereum's extensive transparent governance records—thousands of hours of All Core Dev calls, documented EIP processes, and archived discussions—provide a unique training dataset for LLMs, creating a moat that competitors would struggle to replicate.
Key infrastructure components are also part of the vision. This includes real-time AI moderation for developer calls, an expanded Forkcast.org platform to broadcast governance processes adaptively, and the creation of a cross-client core dev team focused solely on an AI-generated reference client. The ultimate goal is to enhance governance scalability and speed while maintaining decentralization, with AI acting as an amplifying tool rather than a replacement for human judgment.
