Ethereum co-founder Vitalik Buterin identified the limits to human attention as the core problem facing decentralized autonomous organizations (DAOs) and democratic governance systems.
Summary
- Buterin says limited human attention is the main governance flaw of DAOs.
- Personal AI agents could vote based on user preferences and context.
- Suggestion markets and MPC can improve privacy and decisions.
Buterin argued on X that participants must make thousands of decisions across multiple domains of expertise, without enough time or skill to properly evaluate them.
The usual solution of delegation creates powerlessness where a small group controls decision-making while supporters have no influence after clicking the delegation button.
Buterin proposed personal large language models as a solution to the attention problem and shared four approaches. Agents for personal governance, agents for public conversations, suggestion markets, and privacy-preserving multi-party computation for sensitive decisions.
You might also like: Zcash price drops as Ethereum plans stealth addresses and ZK privacy features
Personal LLMs can vote based on preferences
Agents in the realm of personal rule would make all necessary votes based on preferences derived from personal writing, conversation history, and direct statements.
When faced with uncertainty about voting preferences and an issue is important, the agent should ask the user directly, providing all relevant context.
“AI becomes the government” is dystopian: it leads to sloppy behavior when AI is weak, and to maximizing mischief once AI becomes strong. But AI can be used properly and can push the boundaries of democratic/decentralized forms of government.
The core problem with democratic/…
— vitalik.eth (@VitalikButerin) February 21, 2026
Public Conversation Agents collected information from many participants before giving each person or their LLM a chance to respond.
The system would summarize individual views, convert them into shareable formats without exposing private information, and identify similarities between inputs, similar to LLM-enhanced Polis systems.
Buterin noted that good decisions cannot come from “a linear process of taking people’s views based only on their own information, and averaging them (even quadratically).” “Processes must first collect collective information and then enable informed responses.
You might also like: Tether will phase out the offshore yuan stablecoin CNH₮ due to low demand
Suggestion markets could produce high-quality proposals
Governance mechanisms that value high-quality inputs could implement prediction markets where everyone submits proposals while AI agents bet on tokens. When the mechanism accepts the input, it pays out to token holders.
The approach applies to proposals, arguments, or conversation units that the system passes to participants. The market structure creates financial incentives to bring out valuable contributions.
Decentralized governance fails when important decisions require secret information, Buterin argued. Organizations generally handle hostile conflicts, internal disputes, and compensation decisions by appointing individuals with great power.
Multi-party computation using trusted execution environments could include input from many people without compromising privacy.
“You place your personal LLM in a black box, the LLM sees private information, makes a judgment based on it and only displays that judgment,” Buterin explains.
Privacy protection becomes important as participants submit larger inputs with more personal information. Anonymity needs ‘zero-knowledge proofs’, which Buterin believes should be built into all governance tools.