Vitalik Buterin says the biggest problem facing DAOs isn’t technology — it’s human attention.
The Ethereum co-founder wrote on X that people simply don’t have enough time or expertise to properly review the thousands of decisions that decentralized autonomous organizations (DAOs) ask them to vote on.
The real issue: attention
According to Buterin, DAO participants are expected to vote on complex topics across many areas. Most people don’t have the time to study every proposal carefully.
The usual fix is delegation — handing voting power to someone else. But Buterin says that creates another problem. A small group ends up controlling decisions, while everyone else loses real influence after clicking the “delegate” button.
His solution: personal AI agents
Buterin suggested using personal large language models (LLMs) to help solve the attention problem.
He outlined four ideas:
- Personal governance agents
These AI agents would vote on a user’s behalf based on their past writing, conversations, and stated preferences.
If the AI is unsure about how to vote on an important issue, it would ask the user directly and provide full context. - Public conversation agents
These systems would gather and summarize views from many participants before asking for responses.
The goal is to combine shared knowledge first — instead of simply averaging uninformed opinions. - Suggestion markets
This idea uses prediction-style markets where people submit proposals and AI agents “bet” on which ideas are valuable.
If a proposal is accepted, token holders receive rewards. This would create financial incentives for high-quality ideas. - Privacy-preserving systems
For sensitive decisions, Buterin suggested using multi-party computation and secure environments.
In this setup, a personal AI could review private information and make a judgment — but only output the final decision, without revealing personal data.
Privacy matters more with AI
Buterin stressed that as governance systems collect more personal input, privacy becomes critical. He said zero-knowledge proofs and strong anonymity tools should be built into governance platforms from the start.
