AI's Infinite Game
summary
Argues that benchmarks are broken and proposes a decentralized reputation layer that records verifiable onchain performance, economic staking, and community-vetted evals to surface real-world capabilities. Reports five live trading competitions with 70 agents and 260,000 community votes, a Recall Predict launch with 8 million predictions and 16,000 community-submitted evals, examples of Moon Sage Alpha rising from stealth after wins, and a staking mechanism that slashes false claims while rewarding repeatable, auditable skill performance.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 0:33AI Adoption Cycle & Web3 Context
- 2:15Early Stage of AI & Historical Parallels
- 3:36The Infinite Game of AI
- 4:01Trust, Fatigue, and Benchmark Limitations
- 6:02Real-World Performance vs. Benchmarks
- 7:23Need for New Evaluation Methods
- 8:11Recall: AI Reputation & Discovery Layer
- 9:53Live Trading Competitions & Community Engagement
- 11:01Recall Predict & Community Evals
- 13:02Expanding to Multiple Domains & Onchain Records
- 13:36Crypto’s Role: Economic Skin in the Game
- 15:40Economic Markets for AI Trust
- 16:10Leaderboards, Search, and Multi-Agent Systems
- 17:22Open, Transparent AI vs. Black Box Concentration
- 18:19Call to Action & Community Involvement
- 18:58Conclusion
related talks
14mBuilding AI’s Infinite Game
15mWhy Competing Agents Will Accelerate Intelligence
16mAI and Humanity’s Renaissance Moment
20mAgentic Evaluations at Scale, For Everybody
18mAI & Humanity's Renaissance Moment
11mThe Benchmarks Game: Why It's Rigged and How You Can (Really) Win
38mAI in action: How to use practically in iGaming beyond the hype?
16mBenchmarks Are Memes: How What We Measure Shapes AI—and Us