Raoul Pal: The Journey Man The AI Intelligence Tsunami Is Here
71 snips
Feb 26, 2026 
Emad Mostaque, founder and CEO in the AI space known for building Intelligent Internet, explains why 2026 feels like a rapid AI takeoff. He covers agent proliferation, collapsing intelligence costs, and agents automating browsers, workflows and payments. They discuss energy-driven acceleration, world models beyond language, persistent memory, and how agents reshape finance, creative production and software moats.
AI Snips
Chapters
Transcript
Episode notes
Agents Collapse Friction And Costs
- Agents collapse friction in tasks by running continuously and executing complex workflows from chat to web automation.
- Talos' 15,000 tokens/sec demo and Sam Altman's forecast of 100x cost drops mean an average person's annual words could cost ~$$1, making intelligent agents effectively free.
AIs Will Communicate Beyond Human Language
- Continuous learning plus shared priors lets AIs avoid repeated mistakes and compress communication far beyond human limits.
- Emad and Raoul note models will shift from English to machine-native protocols and joint embedding/world models for far greater efficiency.
Persist Memory To Enable Compounding Agents
- Store and organize AI memory externally using markdown, KBs, or QMD to enable persistent, compounding agent intelligence.
- Toby Lütke's QMD example shows fast lookups hooked into Claude-style agents for near-perfect recall.