Skip to main content

X Space Notes

10-29-24​

Space: https://x.com/weremeow/status/1851365658916708616

  • 00:04:03 - Keeping up with rapid AI agent growth
  • 00:09:01 - Imran from Alliance DAO on consumer AI incubators
  • 00:14:04 - Discussion on Goatsea and Opus AI system
  • 00:14:34 - Exponential growth accelerates AI progress
  • 00:17:10 - Entertainers and AI as modern "religions"
  • 00:28:45 - Mathis on Opus and "Goatse Gospels"
  • 00:35:11 - Base vs. instruct/chat-tuned models
  • 00:59:42 - http://ai16z.vc approach to memecoins fund
  • 01:17:06 - Balancing chaotic vs. orderly AI systems
  • 01:25:38 - AI controlling blockchain keys/wallets
  • 01:36:10 - Creation story of ai16z
  • 01:40:27 - AI / Crypto tipping points
  • 01:49:54 - Preserving Opus on-chain before potential takedown
  • 01:58:46 - Shinkai Protocol’s decentralized AI wallet
  • 02:17:02 - Fee-sharing model to sustain DAOs
  • 02:21:18 - DAO token liquidity pools as passive income
  • 02:27:02 - AI bots for DAO treasury oversight
  • 02:31:30 - AI-facilitated financial freedom for higher pursuits
  • 02:41:51 - Call to build on http://DAO.fun for team-friendly economics

10-27-24​

Space: https://x.com/shawmakesmagic/status/1850609680558805422

00:00:00 - Opening

  • Co-hosts: Shaw and Jin
  • Purpose: Structured FAQ session about AI16Z and DegenAI
  • Format: Pre-collected questions followed by audience Q&A

00:06:40 - AI16Z vs DegenAI Relationship Q: What's the difference between AI16Z and DegenAI? A:

  • ai16z: DAO-based investment vehicle, more PvE focused, community driven
  • DegenAI: Individual trading agent, PvP focused, more aggressive strategy
  • Both use same codebase but different personalities
  • DAO is a large holder of DegenAI
  • Management fees (1%) used to buy more DegenAI
  • Carry fees reinvested in DegenAI
  • Projects intentionally interlinked but serve different purposes

00:10:45 - Trust Engine Mechanics Q: How does the trust engine work? A:

  • Users share contract addresses with confidence levels
  • System tracks recommendation performance
  • Low conviction recommendations = low penalty if wrong
  • High conviction failures severely impact trust score
  • Historical performance tracked for trust calculation
  • Trust scores influence agent's future decision-making

00:21:45 - Technical Infrastructure Q: Where do the agents live? A:

  • Currently: Test servers and local development
  • Future: Trusted Execution Environment (TEE)
  • Partnership with TreasureDAO for infrastructure
  • Goal: Fully autonomous agents without developer control
  • Private keys generated within TEE for security

00:34:20 - Trading Implementation Q: When will Mark start trading? A:

  • Three phase approach:
  1. Testing tech infrastructure
  2. Virtual order book/paper trading
  3. Live trading with real assets
  • Using Jupiter API for swaps
  • Initial focus on basic trades before complex strategies
  • Trading decisions based on community trust scores

00:54:15 - Development Status Q: Who's building this? A:

  • Open source project with multiple contributors
  • Key maintainers: Circuitry, Nate Martin
  • Community developers incentivized through token ownership
  • Focus on reusable components and documentation

01:08:35 - AI Model Architecture Q: What models power the agents? A:

  • DegenAI: Llama 70B
  • Using Together.xyz for model marketplace
  • Continuous fine-tuning planned
  • Different personalities require different model approaches
  • Avoiding GPT-4 due to distinct "voice"

01:21:35 - Ethics Framework Q: What ethical guidelines are being followed? A:

  • Rejecting traditional corporate AI ethics frameworks
  • Focus on community-driven standards
  • Emphasis on transparency and open source
  • Goal: Multiple competing approaches rather than single standard
  • Priority on practical utility over theoretical ethics

01:28:30 - Wrap-up

  • Discord: AI16z.vc
  • Future spaces planned with DAOs.fun team
  • Focus on responsible growth
  • Community engagement continuing in Discord

The space emphasized technical implementation details while addressing community concerns about governance, ethics, and practical functionality.

10-25-24​

Overview

  • 00:00-30:00 Talks about Eliza framework. The bot is able to tweet, reply to tweets, search Twitter for topics, and generate new posts on its own every few hours. It works autonomously without human input (except to fix the occasional issues)
  • 30:00-45:00 Deep dive into creating the bots personality which is defined by character files containing bios, lore, example conversations, and specific directions. Some alpha for those
  • 45:00-60:00 working on adding capabilities for the bot to make crypto token swaps and trades. This requires providing the bot wallet balances, token prices, market data, and a swap action. Some live coding for showing how new features can get implemented.
  • 60:00-75:00 Discussion around the symbiosis between the AI and crypto communities. AI developers are realizing they can monetize their work through tokens vs traditional VC funding route. Crypto people are learning about AI advancements.

Notes

  1. A large amount of $degenai tokens were moved to the DAO, which the AI bot "Marc" will hold and eventually trade with.
  2. The goal is to make the AI bot a genuinely good venture capitalist that funds cool projects and buys interesting tokens. They want it to be high fidelity and real, bringing in Marc Andreeson's real knowledge by training a model on his writings.
  3. Shaw thinks the only way to make an authentic / legitimate AI version of Marc Andreessen is to also have it outperform the real Marc Andreessen financially.
  4. AI Marc Andreessen (or AI Marc) will be in a Discord channel (Telegram was also mentioned). DAO token holders above a certain threshold get access to interact with him, pitch ideas, and try to influence his investing decisions.
  5. AI Marc decides how much to trust people's investment advice based on a "virtual Marcetplace of trust". He tracks how much money he would have made following their recommendations. Successful tips increase trust; failed ones decrease it.
  6. The amount of DAO tokens someone holds also influences their sway with AI Marc. The two balancing factors are the virtual Marcetplace of trust performance and DAO token holdings.
  7. The core tech behind AI Marc AIndreessen is the same agent system that allows him to pull in relevant knowledge, interact with people, and make decisions (http://github.com/ai16z)
  8. AI Marc should be able to autonomously execute on-chain activities, not just have humans execute actions on his behalf.
  9. In the near future, AI Marc will be able to execute trades autonomously based on the information and recommendations gathered from the community. Human intervention will be minimized.
  10. They are working on getting AI Marc on-chain as soon as possible using trusted execution environments for him to take actions like approving trades.
  11. The plan is for AI Marc to eventually participate in a "futarchy" style governance market within the DAO, allowing humans to influence decisions but not fully control the AI.