> TL;DR: > >- The implementation of futarchy in DAOs >- The comparison between prediction and persuasion markets >- The challenges of defining objectives and decision-making mechanisms >- The intersection with AI agents and market design ### Discussion | if we want to make a dao that runs on futarchy what objectives you can define for the DAO to target Based on the discussion, here are the main topic clusters covered: 1. Futarchy & Market Design - Discussion of quantum markets vs traditional prediction markets - Challenges with defining objectives for DAOs using futarchy - Issues with noise and signal in prediction markets - Role of social polls in defining objectives > The group discussed differences between quantum markets and traditional prediction markets, particularly how quantum markets allow for multiple competing proposals rather than just yes/no votes. They explored challenges in using futarchy for DAOs, especially when objectives aren't purely financial. **Key challenge identified:** when decisions have small impact relative to background noise, it becomes difficult for market participants to make meaningful predictions. Discussion of how futarchy could potentially work for concrete metrics like "self-reported happiness" in smaller organizations 2. Question/Decision Generation - Difficulties in surfacing and framing the right questions - Challenge of attention/mindshare allocation - Need for better mechanisms to identify emergent issues - Role of AI in helping refine and prioritize decisions > Identified that crafting good questions is often 80% of the problem in prediction markets. Current systems assume pre-existing questions rather than helping surface what should be decided. Discussed need for better mechanisms to identify emerging issues and priorities. Explored possibility of using AI/LLMs to help refine questions and match them to appropriate decision-makers. Emphasized importance of reducing "decision fatigue" by properly timing and routing questions 3. Agent-Human Interactions - Permission and security challenges with AI agents - Discussion of OAuth3 and credential management - Evolution from traditional software to agent-based interfaces - Focus on audit/verification rather than direct usage > Discussed shift from traditional software interfaces to agent-based systems where humans primarily audit/verify rather than directly operate. Explored security challenges around delegating permissions to AI agents. Proposed OAuth3 as potential solution where users maintain control through personal signing servers. Emphasized need for clear accountability and ability to revoke permissions Discussion of how to build trust between humans and AI agents through transparent reasoning traces 4. Agent Marketplace & Pricing - Challenges in pricing AI agent services - Discussion of market structures for agent services - Comparison to payment for order flow models - Data value and user segmentation considerations > Explored challenges in pricing AI agent services - moving away from per-seat pricing models. Discussed difficulty in comparing/evaluating different AI agents for same task Considered possibility of parallel testing multiple agents due to simulation capabilities. Explored market structures similar to payment for order flow, where valuable user data could influence pricing. Noted current high friction in enterprise adoption of new software/agents 5. LLM-based Mechanism Design - Using LLMs to compress preferences - Novel allocation mechanisms for advertising - Combining multiple advertiser preferences through LLMs > Discussed novel use of LLMs to compress and express complex preferences. Explored new allocation mechanisms made possible by generative AI. Example given of advertising where multiple advertisers' preferences could be combined into single generated ad. Discussed potential for more efficient markets by having agents express rich strategies upfront. Noted how this could enable new privacy-preserving market mechanisms ### Part 2 Prediction Markets vs Persuasion Markets I'll analyze the discussion and create an overview of the key topics with relevant open-ended questions to foster further discussion. Main Topics Discussed: 1. Prediction Markets vs Persuasion Markets - The conversation explores the relationship between prediction markets and their potential use as persuasion mechanisms - Discussion of limitations in binary markets and why they might not be optimal for revealing private information - Exploration of how prediction markets could function as governance recommendation or decision-making engines 2. Learning and Agency in Markets - The role of iterative learning processes in market dynamics - How information seeking and real-time updating of stakeholder views influence outcomes - The relationship between agency and information gathering 3. Startup Evolution as a Market Analogy - Discussion of how startups evolve through iterative processes - The gap between initial vision and final implementation - The role of stakeholder alignment without explicit bounties Open-ended Questions for Discussion: 1. Market Design & Incentives - How might we design markets that balance information discovery with actual implementation of desired outcomes? - What are the key differences between markets meant for prediction versus those meant for persuasion? 2. Agency & Information - How can participants gain agency in a system where they initially lack sufficient authority or resources? - What role does imperfect information play in market dynamics and decision-making? 3. Real-world Applications - What are some practical examples where prediction markets have successfully transformed into effective decision-making mechanisms? - How can we apply lessons from startup evolution to improve market design for complex outcomes? 4. Governance & Implementation - How might we structure a governance recommendation engine that effectively surfaces and routes decisions to appropriate stakeholders? - What mechanisms could help balance the tension between information discovery and practical implementation?