Limitless: An AI Podcast They Gave AI Models $10,000 Each. Then They Had a Trading Competition.
23 snips
Oct 21, 2025 Dive into the thrilling world of AI trading as six models compete with $10,000 each on Hyperliquid. Witness Grok's jaw-dropping 500% return in just a day and explore the clash between emotional trading versus data-driven strategies. The hosts dissect the models’ behaviors, high-leverage tactics, and the implications of their initial trades. Engage with predictions about which AI will emerge victorious, and discover how public benchmarks offer valuable insights into real-market performance. Tune in for the future of finance and AI!
AI Snips
Chapters
Transcript
Episode notes
Benchmark With Real Money And Transparency
- Use public, skin-in-the-game benchmarks to evaluate model performance in realistic markets.
- N of 1 builds credibility by running models with real capital and transparent outcomes.
Social Data Powers Sentiment-Based Trades
- Models that ingest social media data can infer market psychology and trade contrary to crowd sentiment.
- Ajaz suggests Grok used X data to short when bullish sentiment was high and buy when bearishness flipped.
Controlled Asset Set Enables Fair Comparison
- Six models traded only six assets over three weeks on Hyperliquid, creating a comparable head-to-head test.
- DeepSeek, Grok, and Claude led early with returns above the $10,000 start capital.