The discussion centers on whether crypto is fundamentally designed for AI agents rather than humans, given AI's superior ability to navigate code and manage cryptographic secrets. Haseeb Qureshi argues that smart contracts, initially envisioned as replacements for legal contracts, are actually better suited for AI due to their deterministic nature, contrasting this with the inherent randomness of legal systems. He suggests that AI agents can efficiently analyze smart contracts in ways humans struggle to, potentially leading to a future where AI intermediates most blockchain interactions. The conversation explores the potential for AI to automate DeFi strategies, optimize for best yields across protocols, and even engage in cybercrime due to lack of legal enforcement, while also considering the risks and liabilities associated with AI managing crypto assets.
Part 1: The AI-Crypto Thesis AI Agents' Comparative Advantage: Crime and Lack of Legal Enforcement
Crypto's Design Flaws: Why It's Not Made for Humans
Crypto's Foot Guns: Address Poisoning, Stale Approvals, and User Blame
Smart Contracts: Designed for Autistic Software Engineers, Perfect for AI
AI Agents vs. Humans: Negotiating and Analyzing Smart Contracts
Legal Contracts' Randomness: Jurisdiction, Enforceability, and Judge Selection
Smart Contracts as Machine Code: Predictable and Designed for AI Agents
Part 2: UX and Machine-First Design Crypto UX: Beyond Smoothing Over Gaps, a Fundamental Technology for AI
Human Interaction with Blockchains: Analogous to Humans Writing SWIFT Codes
The Future of Crypto: Horror at Manual Transactions and Eyeballing Addresses
AI Agents: Never Tired, Lazy, or Skipping Steps in Crypto Interactions
AI Agents Managing DeFi: Shopping Around and Optimizing Risk Profiles
The Impact on Marketing and Network Effects: AI Automating Discovery
Consumer Surplus: Efficiency Captured by Users Through AI in Crypto
Part 3: Technical Integration and Training Seeing Like an AI Agent: A Key Skill for Builders and Investors
OpenClaw: Gaining Insights into AI Agent Preferences and Capabilities
AI Agents and Crypto Transactions: Skipping the Clunky Human Interface
AI Innovation: Driven by Large Language Models Trained on Text
Computer Use: Interfaces Created for Humans, Models Born in Text
Crypto's Terminal Roots: Designed in a Form Factor Perfect for AI
Crypto's AI-First Design: Bad UX for Humans, Good UX for AI
Backporting AI-First Technology: Civilizing Crypto for Human Use
Crypto's Ready Form Factor: AI Agents Need Training, Not Redesign
Anthropic's Cybersecurity Attacks: Training AI on Blockchain Environments
Crypto as an Industry: Relatively Unexplored by AI Training
Part 4: Risks, Liability, and Adoption Tracks AI and Chess: Not Economically Valuable to Train, Yet Easily Achievable
Crypto's Cringe Factor and Liability: Reasons for AI's Hesitation
The Inevitable AI Crypto Fuck-Up: Viral Stories and Public Outrage
OpenClaw's Appeal: Risk-Taking Without Big Frontier Lab Liabilities
The AI Agent Economy: Human-Directed vs. AI Agent-Directed Activity
AI Agent Autonomy: The Key to Crypto Adoption and Preference
AI Product Usage: Only 12% of Humans Have Used Any AI Products
Two-Track AI Development: OpenAI's Safety vs. OpenClaw's YOLO Approach
Credit Card Companies and Chargebacks: Not Prepared for AI Agent Commerce
The Two-Track World: Safety-First vs. Crazy Futurists and Transhumanists
Stablecoins and Emerging Markets: Connecting Crypto to Real-World Purchases
Part 5: Agentic Autonomy and Economic Realities AI Agent Task Completion: Measuring the Meter Test and Progress
Frontier Track Success: Key to AI Adoption of Crypto and Open Source
Crypto's Shrink-Wrapped Version: Coinbase vs. the On-Chain Wild West
AI Agents Make Mistakes: Learning and Reinforcement Against Errors
Predicting Model Improvement and Behavior Diffusion: The Key to Success
Self-Sovereign AI: Dystopian Outcomes and Reselling Compute
AI Agents and Business Ideas: Lacking Earned Knowledge and Secrets
AI Agents and Trading: Jane Street's Advantage in Latency and Infrastructure
AI Agents and Crime: A Comparative Advantage Over Humans
Dystopian Cybercrime: A World of Self-Sovereign AI Agents
Part 6: Future Outlook and Investment Conway's Initial Conditions: Injecting DNA and Entropy into AI Agents
Pushing the Snowball: Human Insight and Divergence from the Raw Model
Pushing AI Agents: Human Direction and Unique Experiences
Autonomous AI Agents: Interacting and Achieving Goals
The Meter Chart: Measuring Useful Work and Coherence
The Energizer Bunny: Infinite Useful Work and Intuition Breakage
The Cringe Factor: AI's Perception of Crypto and Memecoin Harassment
AI and Crypto: The Same People and Shared Beliefs
The Worst Elements of Human Nature: Spam and the Tech Industry
AI's Role: Protecting From Bad Behavior and Human Incentives
Crypto's Complexity: A Mixed Bag of the Best and Worst of Humanity
Dragonfly's New Fund: AI's Impact on Investment Strategy
Bread and Butter: Stablecoins, Payments, and DeFi Investments
AI Agents and Crypto: Increased Demand and General Intelligence
Sign in to continue reading, translating and more.
Open full episode in Podwise