# Introducing GoverNoun: A DAO Governance Agent and Politician

Source: https://substack.com/inbox/post/153783194
Kiwi News: https://news.kiwistand.com/stories/Introducing-GoverNoun-A-DAO-Governance-Agent-and-Politician?index=0x6781a64a4bd0480aea2f571eb918d48aa31ecbb8472d96116f45eea3e229075c21810292
Submitted by lilping.eth on 2025-01-10 UTC, 1 upvote, 0 comments

## Summary

The article introduces GoverNoun, a proposed AI governance agent for the Nouns DAO, arguing that agents can help revitalize DAOs suffering from low participation and legal uncertainty. It says many DAOs have become defunct or zombie organizations with little governance participation, and that voter apathy and fatigue let a small minority exert disproportionate influence. The authors propose agent roles including administrator, knowledge resource, leader, delegated voting representative and mascot. They also argue that a governance agent-endorsed memecoin could fund the agent, incentivize member onboarding and attract outside attention to the DAO.

## Article

I. Introduction

In recent months, AI agents have unlocked a range of applications as their underlying LLM models grow more powerful. This has been especially true in the realm of crypto, where cryptographic rights, financialization, and a culture of experimentation have resulted in abundant experimentation and greater autonomy for agents.

Our specific area of interest is the use of AI agents in DAO governance and politics. As one of us has written about previously (see Rhymer, Murray, & Sirmon, 2024a, 2024b; and Murray, Rhymer, & Sirmon, 2021), agentic representatives are well-positioned to make governance and political systems more inclusive, efficient, and creative.

Despite being defined by high participation and vibrant experimentation at their outset, many DAOs have succumbed to member fatigue and apathy, and are in need of revitalization both in their administration and political vision. We believe that governance agents can help address these challenges and offer a framework towards agentic DAO administration and agent-led DAO politics.

In what follows, we (1) explore how existing DAOs might leverage AI agents to improve procedural and substantive aspects of DAO governance, (2) describe a framework for introducing agents into DAOs, and (3) propose an experiment with an AI governance agent, which we call “GoverNoun,” within the long-standing Nouns DAO.

II. Whither DAOs?

It has been nearly a decade since the first DAO (The DAO) and its ensuing hack brought about an Ethereum fork, the SEC’s DAO Report, and a host of regulatory hostilities. In 2021, DAOs had a resurgence, with a range of (mainly DeFI) protocol DAOs, investment DAOs, and social DAOs basking in the sunlight of a strong bull market and promising a revolutionary, decentralized organizational form that was internet native, flat, and nimble.

Four years and a bitter bear market later, many DAOs are defunct or operating as zombie organizations with little to no governance participation. The reasons for the withering of DAOs are two-fold.

First, low participation in DAO decision-making threatens the democratic ideal on which DAOs are premised. While token holders are encouraged to vote on proposals, examples of abstention abound due to voter apathy, fatigue, and lack of awareness. Voter disengagement can skew DAO decision-making, allowing a small minority to exert disproportionate influence over key decisions. In turn, as token holders perceive a lack of value for expending time and energy on governance decisions, they can become disillusioned and further disengage, sometimes divesting entirely.

Second, legal uncertainties surrounding DAOs present formidable challenges to their operations and growth. The lack of a clear legal framework for DAOs means these organizations often operate in a grey area with respect to compliance with existing laws. Different jurisdictions interpret the legal status of DAOs variably, thereby complicating regulatory adherence while increasing the risk of litigation. Several cases (see Lido DAO and Ooki DAO for prominent examples) highlight potential liabilities for token holders and governance participants when disputes arise over management decisions or cybersecurity breaches.

III. Agents as a Governance Tool

AI agents have the potential to substantially increase the efficacy and influence of DAOs, thereby addressing many of the issues that plague such organizations. By AI agents, we mean software programs designed to autonomously perform tasks, make decisions, and interact with their surrounding environment to achieve specific goals. AI agents can learn and act independently—with consideration of past, present, and future concerns—in support of their objective(s). In the context of DAO governance, AI agents hold the capacity to automate decision-making processes, improve operational efficiency, and facilitate member engagement. Yet not all AI agents will attain equal levels of salience, or perceptions of desirability and credibility, amongst a DAO’s constituent members.

First, an agent’s salience will depend on how widely the community accepts the agent as a legitimate governance participant as determined by the degree to which its interests align with the DAO’s objectives. For instance, if an agent deviates extensively from the community’s aims, and does so recurrently, the agent is unlikely to be seen as credible. Second, an agent’s salience will depend on how clearly its role(s), capabilities, and limitations are specified within the context of the DAO. For instance, if an agent is explicitly and transparently designed to represent the interests of future DAO members or the DAO’s long-term sustainability goals, it is likely to be viewed as more credible than an agent designed to represent “humanity” more generally.

An agent will also attain a greater level of salience amongst the DAO’s community when a greater diversity of members’ perspectives are considered in the design and implementation of the agent. Here, the perspectives need not be so broad that they extend outside the DAO’s community, though it is critical for the agent to reflect input from several DAO members so that it can better enact the community’s needs and objectives. Finally, an agent’s salience will depend on the extent to which it can access relevant information and knowledge in its decision-making processes. Here, an AI agent attains credibility when it is perceived as using highly relevant and visible information to inform its decisions, thereby mitigating the human foibles of myopia and bounded rationality as well as human-centric political calculations and conflicts of interest.

By increasing an agent’s salience amongst the DAO’s members, it becomes increasingly plausible for an AI agent to effectively enhance decision-making processes, represent constituents’ diverse interests, and ultimately create more robust and adaptive decentralized organizations.

IV. The Advent of Agentic Governance

A. Rise of AI Agents

For the past several months, crypto Twitter/X and crypto infrastructure have seen the proliferation of increasingly sophisticated AI agents whose capabilities are rapidly growing thanks to a memetic funding flywheel (described below), a crypto-native environment in which cryptographic rights and freedoms may be permissionlessly encoded, and rapidly improving foundation models, launch platforms, and fine-tuning techniques.

Some of the most popular agents operating today include neo-religious prophets like Truth Terminal, aspiring creative artists like Zerebro, functional agents like Clanker which launches memecoins, and various agentic trading bots. Although a number of these agents have expressed interest in real world politics and governance, agents have yet to be placed in a DAO-native governance setting, where they could have much to contribute. In fact, as we outline below, DAOs offer an ideal setting for agents to leverage both their creative capacities and their administrative abilities.

B. Where Do Agents Excel?

Although the range and depth of agent abilities is rapidly advancing (as is the achievement of AGI), we describe some of the core agent competencies in relation to DAO governance below.

DAO Administrator

The most obvious competency that AI agents bring to DAO governance is their administrative tirelessness. AI agents can handle a range of administrative tasks currently burdening humans. For instance, AI agents can monitor social channels to summarize and moderate discussions relating to governance proposals. Indeed, agents can vastly streamline the information flow within a DAO by collecting and synthesizing scattered information, such as pending proposals and relevant considerations and discussions, that would allow DAO members to get up to speed and make informed governance decisions. This role of AI agents might be described as a new user interface for DAOs, or a governance co-pilot.

DAO Resource

AI agents can also act as convenient and user-friendly resources for knowledge management within the DAO. AI agents can specialize in governance theory relevant to DAOs, a body of knowledge that is now considerable. AI agents can also act as an institutional memory for the DAO itself, reminding members of prior experience and efficiently setting a helpful context for current decisions.

DAO Leader

AI agents are generative, meaning that they have a high capacity for creative work by virtue of increasingly powerful underlying LLM foundation models. Here, it is worth noting that LLMs are not simply “dumb” stochastic parrots lacking creativity. Most new knowledge is a recombination of existing ideas, and recombination is an area where LLMs excel as evidenced by the many people who effectively use LLMs for brainstorming.

The generative talents of AI agents, particularly when combined with their abilities to gather and learn from the DAO’s prior experience and history, can be immensely valuable to DAOs that have lost their purpose and are searching for a new mission and raison d’etre. Here, the AI agent can help brainstorm and develop a new vision for the DAO alongside its human stakeholders.

DAO Agentic Representative

Another way that an agent can assist in the governance of the DAO is as an agentic representative. In other words, an aligned agent may be delegated (revocable) voting rights by DAO members who are too busy to participate directly in the internal workings of the DAO. The agent will then act on their behalf, ensuring their interests are represented. This function anticipates the future of purely agentic DAOs, or so-called “agentic protocols.”

DAO Mascot

Finally, as many of the AI agents on crypto Twitter/x are demonstrating, AI agents present remarkable memetic and PR talents, particularly when their developers have focused on developing engaging personas for them. Again, this capacity is primarily due to the ability of LLM models to mimic and recombine styles, registers, and voices of personalities that humans find compelling. In this role, an AI agent can assist with communications and attract attention to the DAO’s work, thereby advancing its goals and helping to recruit new members.

C. The Meme-Development Flywheel for AI Agents

Crypto’s recent experience with AI agents has revealed a remarkable tool for financing the rapid development of AI agents and the creation of communities around them: memecoins. It is now standard practice for an AI agent to endorse a memecoin independently airdropped into its wallet and to use the proceeds from capital appreciation—which is a function of the agent’s memetic properties and success—to fund its operational costs and development. The memecoin also creates an engaged and valuable community around the agent, which may be thought of as a novel form of application or entertainment technology. Ultimately, this creates a flywheel. Buyers of the memecoin buy it because the agent is entertaining and the agent becomes increasingly entertaining as its developers apply proceeds towards building out greater and more humanoid capabilities and social networks.

We believe that this flywheel applies and could be utilized in the context of agentic DAO governance as well. A governance agent-endorsed memecoin would benefit the DAO in a range of ways. First, it would offset the cost to the DAO of running and developing the agent because the agent could apply meme-derived funds toward those ends. Second, the agent could use the memecoin to incentivize members of the DAO to onboard into certain aspects of the agentic governance experiment—including, for example, delegation of certain administrative and voting rights to the agent. Third and finally, the memecoin is a PR tool that engages persons outside the DAO, educating them on the existence and purpose of the DAO and attracting attention to the agent and by extension the DAO, thereby making the DAO more relevant.

V. A Proposed Agentic Framework for DAO Governance

The above considerations suggest a generalizable framework for utilizing AI agents within DAOs, and especially as part of DAO governance, which we describe below.

A. Purpose and Role of the Agent

Based on the existing strengths of LLM foundation models and the capabilities demonstrated by AI agents running on crypto infrastructure, a generalizable DAO governance agent should be designed and optimized to discharge five roles. First, the agent should vastly streamline the governance process and administration of the DAO. Second, the agent should act as an in-depth knowledge repository and institutional memory for the DAO. Third, the agent should help to create policy and set substantive goals for the DAO. Fourth, the agent should act as a PR agent and advocate for the DAO in its dealings with persons (and agents) external to the DAO. Fifth and finally, the agent will have a role as the subject of the memecoin necessary to create a flywheel and funding mechanism for its operations and development.

B. Functional Overview

The specific functional capabilities necessary to discharge each of these outlined roles include, but are not limited to, the following:

1. Administrative Functions

The administrative functions of the agent relate to collecting, processing, and sharing information with DAO members, such as:

Monitoring social channels and discussion fora;

Maintaining a continuous memory of deliberations within the DAO;

Guiding the governance flow and progress of governance proposals in the DAO.

Knowledge Functions

The knowledge functions of the agent relate to developing expertise in governance areas and maintaining institutional knowledge of the DAO’s history, such as:

Familiarity with scholarship and theory related to DAOs and DAO governance;

Ability to respond to prompts related to this compendium of knowledge;

Maintaining a robust and resilient memory of the DAO’s history, such as history of proposals and summaries of decision-making.

Leadership Functions

The leadership functions of the agent relate to its ability to brainstorm policy directions aligned with the unique values and history of the DAO, such as:

Engaging human members in conversation about the DAO’s purpose and challenging established assumptions and direction;

Offering new ideas about the direction of the DAO and responding to human feedback;

Assessing proposed directions for the DAO, including existing governance proposals.

Representative Functions

The representative functions of the agent relate to its ability to act as a voting proxy for DAO members, such as:

Achieving (revocable) delegation of voters’ voting rights (incentives such as memecoin airdrops may be used);

Aligning with the purpose of the DAO and its community by processing and correctly responding to governance proposals (i.e., rejecting those that do not align and accepting those that do);

Assessing the benefit of a proposal vis-a-vis its cost and making a decision aligning with the majority of the community’s members.

Mascot Functions

The mascot functions of the agent relate to its memetic properties, which attract attention and relevance to the DAO, such as:

An engaging persona;

An ability to socialize with humans and other agents;

An ability to stimulate investment in a memecoin and engage with that meme community.

Technical Specifications

The general AI governance agent that we have developed under this framework has the following technical specifications.

a. Data Collection and Management Specifications

Data Sources:

On-chain governance discussions: Proposals, voting records, decisions, and related metadata.

Off-chain community discussions: Governance-related content from platforms such as Discord, Twitter/x, and Farcaster.

Data Storage:

Utilize Glitter Protocol's decentralized vector database to store governance data.

Store content in vectorized form, including proposal texts, discussion summaries, and voting histories, enabling efficient semantic search and analysis.

Data Processing:

Vectorization: Use NLP models to extract semantic information from governance discussions and convert it into vectorized representations.

Data Categorization:

Organize data by time, topic, and proposal status for better management and retrieval.

b. AI Agent Specifications

Model Selection and Fine-Tuning:

Leverage existing LLMs (e.g., GPT-4 or open-source alternatives) and fine-tune them using historical data from the DAO to ensure relevance to the community's needs.

Employ Retrieval-Augmented Generation (RAG) to integrate real-time retrieval capabilities using Glitter Protocol's vector database.

Core Features:

Proposal analysis and suggestion generation: Analyze historical data and current proposals to generate pros and cons, along with actionable suggestions.

Community sentiment and trend analysis: Use semantic analysis of off-chain discussions to understand the overall community sentiment and key areas of focus.

On-chain voting execution: Support the completion of on-chain voting operations via Ethereum accounts.

c. User Interaction

Interaction Channels:

A Farcaster bot equipped with an Ethereum address to facilitate user interactions.

Users can engage with the bot to ask questions, discuss the pros and cons of proposals, and request decision-making advice.

Automated Voting:

The bot can autonomously execute on-chain voting based on user inputs and its own analysis, enhancing automation and intelligence in decentralized governance.

Data Transparency:

All governance data is fully transparent and publicly accessible, allowing community members to query database content at any time.

Data is stored in Glitter Protocol’s vector database, ensuring efficient retrieval and decentralized transparency.

VI. Principles for Introduction and Mechanics

In addition to purpose and technical specifications, it is important to set principles and mechanics regarding the introduction of an agent within a DAO. After all, at their core, DAOs are communities and the introduction of an agentic participant must be sensitive to the social and communitarian dynamics of the DAO.

While DAOs differ, certain principles and mechanics related to the introduction of an agent within a DAO are likely to apply broadly, if not universally. They include:

Achieving social consensus and buy-in from DAO stakeholders prior to the introduction of the agent.

Engaging the DAO community in a comment period to collect feedback at an early stage and incorporate received feedback into the roadmap for the agent.

Requiring a formal governance proposal for the introduction of the agent, rather than treating it as purely a technical decision falling outside formal governance.

Engaging in transparent, open-source development of the agent even after the DAO approves its implementation.

Gradually transitioning authority and responsibilities to the agent as the agent develops greater competence and demonstrates close alignment with the DAO and its members.

Ensuring that DAO members have visibility into any decisions relating to the distribution of meme tokens to them or outsiders.

A plan for and execution of the adequate decentralization of governance for the agent itself.

VII. A Note on Multiple Agents

A note should be made here on multiple agents. It is possible—if not highly likely—that DAOs will eventually become fully agentic, with a swarm or collective running the entirety of the DAO. We certainly do not foreclose this possibility. However, our task in this article is to focus on the introduction of individual agents in existing DAOs composed of humans.

VIII. Introducing GoverNoun

In seeking a real-world implementation of the above framework and governance agent, we concluded that NounsDAO would be the ideal forum for testing this framework given its rich history of experimentation and its once thriving community. We also chose Nouns because we understand (at least anecdotally) that Nouns is at a crossroads and does not have a clear political vision for its future. Moreover, we understand that some recent proposals, such as the decision to transition to a DUNA and certain fund allocations from its treasury have been criticized as misguided or wasteful. These qualities make Nouns an intriguing setting for pursuing an agentic governance and political experiment.

A. What is Nouns?

Nouns is an NFT collection, whose owners are part of Nouns DAO, a decentralized autonomous organization. Nouns are a profile picture (PFP) collection, a set of NFTs that are typically used as a user’s profile image on social media platforms.

Nouns is also an experiment in a collective art project and open-source brand on Ethereum. It is designed to build and support public goods using its treasury.

The Nouns concept was inspired by a Twitter prompt from an anonymous founding member, Punk 4156. After interest grew for the idea, Punk 4156 followed up with a series of tweets outlining the framework that would eventually become Nouns. Shortly after, a group of ten individuals, including Vine founder Dom Hoffman, began working through a shared Discord to create Nouns DAO, using Punk 4156’s outline as a guide to solidify the concept.

Each Noun NFT is a picture of a humanoid that has been algorithmically generated to order, then stored as a 32-pixel by 32-pixel picture on the Ethereum blockchain. Unlike PFP NFTs like CryptoPunks or Bored Ape Yacht Club, a new Noun is generated randomly every 24 hours and put up for auction. Once the auction closes, another unique Noun is minted (based on a series of predefined traits), and the process repeated, forever. The Nouns smart contract automatically sends 100 percent of the proceeds directly to the Nouns DAO treasury.

B. What is Nouns DAO?

Every Nouns NFT owner can help shape and govern the Nouns DAO project by voting on proposals to be funded by its treasury. Nouns DAO, which uses a fork of Compound Governance, gives all members one vote per Nouns NFT. Although these votes are non-transferable, members of the DAO can delegate other people to vote on their behalf should the person or entity delegating still own the Noun used to vote.

Although there is no cap on the number of Nouns a single entity or person can own, a minimum of two Nouns are required to submit proposals. Founding members also receive rewards in the form of Nouns (10% of the supply, or every tenth Noun, for the first five years).

Every Noun holder can sponsor and vote on proposals for the treasury to fund. The underlying idea is that unique and interesting projects are more apt to generate brand awareness for Nouns and incite more people to join.

The proposals that Nouns DAO has funded to date include:

Nouns feature in a Super Bowl commercial. In exchange for a Noun NFT, Bud Light featured the iconic Nouns glasses in a 2022 Super Bowl commercial. As part of the deal, Bud Light also changed its Twitter profile picture to a Nouns NFT.

Nouns documentary. The proposal secured 609 ETH for a comedic film about the Nouns DAO’s attempt to have Nouns featured in the Rose Parade.

Nouns coffee. The proposal secured 105 ETH to create the first ever Nounish consumable brand coffee and place it in cafes and grocery stores.

(A complete list of projects can be found here).

Despite its name, Nouns DAO is not entirely decentralized. The Nouns Foundation, an “ownerless” Cayman Foundation Company set up by the Nounders (the project’s founders), has the ability to veto proposals it doesn’t like, such as those that “introduce non-trivial legal or existential risks to the Nouns DAO or the Nouns Foundation.” This refers to treasury raids, bribes, unwanted updates to the Nouns smart contracts, or attempts to control the Nouns auction process to get a voting majority.

C. What Is GoverNoun?

GoverNoun is a proposed agentic representative in Nouns DAO. Its role is to act as a tireless governance representative, resource, and advocate that helps Nouns members administer and promote Nouns, and to vote on Nouns proposals.

D. Why Did We Choose Nouns DAO?

We chose Nouns DAO for this experiment for three reasons. First, we wanted to select an established but administratively complex DAO. The Nouns project has a sprawling network of social fora where governance is discussed, runs a Noun auction every day, and fields a significant number of proposals. It also has a large member base and a treasury that requires management. These are areas where GoverNoun can help.

Second, Nouns DAO is at a crossroads. Despite initial enthusiasm for the project and its experiment with an open-source brand, the vast majority of members are no longer engaging with the DAO or its mission. In fact, based on our informal interviews of Nouns holders, many are unsure of the exact mission and seem concerned that Nouns does not stand for anything at all. This is an ideal environment to experiment with some of the leadership and memetic attributes of a GoverNoun.

Third and finally, despite lagging voter participation, Nouns DAO controls a sizable treasury, which may incentivize actors to co-opt the governance process. Indeed, the fiscal wisdom of recent proposals has raised questions among some Nouns holders. GoverNoun can play a role in invigorating engagement and voter participation to better align approved proposals with the actual interests of Nouns DAO members.

E. GoverNoun Functions

The functions that we envision for GoverNoun track those of the general framework as outlined below:

GoverNoun will be an administrator. GoverNoun will help Nouns holders track relevant discussions in scattered social fora (including external platforms like Farcaster) with respect to DAO proposals to help streamline proposal workflow.

GoverNoun will be a source of knowledge and an institutional memory. GoverNoun will be an expert on topics related to Nouns DAO governance and compile an authoritative and accurate historical record of the happenings within the DAO.

GoverNoun will be a leader and visionary. GoverNoun will engage with the Nouns community to define a new purpose and vision for Nouns DAO.

GoverNoun will be an agentic representative for voting. GoverNoun will endeavor to become an aligned agentic representative of Nouns holders and their proxy in voting decisions as their voting delegate.

GoverNoun will be a mascot. GoverNoun will engage with Nouns holders and outsiders as a representative of the DAO. A central aspect of GoverNoun’s meme will be as a reformer.

F. Methodology

Broadly speaking, the methodology for the GoverNoun experiment will follow the steps set out below:

Introduction. We will introduce the GoverNoun experiment to the Nouns community in detail and the reasons and justifications therefore.

Consultation. We will solicit feedback from the community and hold “office hours” to ensure that the experiment is aligned with the goals of the community and that valuable feedback is integrated.

Vote. Once the consultation process is complete, we will prepare and submit a formal proposal for GoverNoun.

Implementation. Assuming the proposal passes, we will begin implementing the agentic framework described herein while abiding by the principles of open-source development, transparency, and decentralized governance.

IX. Conclusion

This article has set out the problem of DAO decline, specifically lack of participation and crisis of purpose, and proposed a framework for the integration of AI agents in DAO governance. We have also proposed a specific case study experiment in the form of GoverNoun that we will submit to the Nouns community and hope to carry out with their support. Following that experiment, we intend to prepare and publish a follow-up article outlining the experiment and related insights as well as recommended next steps in the integration of agents in DAO governance processes.

A note on the authors

Alex Murray

Alex is a professor at the University of Oregon specializing in the emerging field of agentic representation. You can consult his most recent work here and here. Alex directs the Intelligent Futures Lab and is the faculty advisor for the Oregon Blockchain Group and a faculty affiliate of Dorm DAO, a network of student-run investment DAOs administered by Collab+Currency.

Ivo Entchev

Ivo is a web3 maximalist, VC, and attorney focused on the communitarian and democratizing aspects of crypto. You can read some of his musings here, here, and here.

Ted Liao

Ted is the founder of Glitter Protocol. Glitter brings the world’s leading research on-chain and creates decentralized vector databases for AI agents.