AI Pod by Wes Roth and Dylan Curious | Artificial Intelligence News and Interviews With Experts GROK 4.20 and the "SOCIETY OF MINDS"
Mar 10, 2026
They unpack Grok 4.20’s beta and its multi-agent “swarm intelligence” along with early performance claims. The conversation explores the society-of-minds idea and research showing collaborative agents can outperform individuals. They debate Anthropic’s legal moves and the geopolitics of AI infrastructure. The show also covers autonomous weapons ethics, agent-run news and GEO-driven SEO experiments.
AI Snips
Chapters
Transcript
Episode notes
Grok 4.20 Adds Swarm Intelligence And Deep Integration
- Grok 4.20 introduces a multi-agent swarm that coordinates specialized agents to solve complex tasks.
- Wes and Dylan note Grok 4.20 runs on a huge Colossus GPU cluster with RealX data, multimodal and Tesla integration, plus personality modes.
Multiple Agents Produce Better Results Than Single Models
- Wisdom-of-the-crowd sampling improves LLM reliability by aggregating many stochastic outputs.
- Researchers evolved this into chatting agents that coordinate roles and outperform single models on tasks like code and Minecraft teamwork.
OpenClaw Society Of Minds Cut A YouTube Tracker From Dollars To Pennies
- Wes built a 'Society of Minds' by connecting multiple big-model APIs so they chat and co-design solutions.
- Example: Claude wrote functioning YouTube tracker code while Gemini suggested using RSS to cut API costs from dollars to pennies.