The Models Trying to Fill the Fable Gap
Summary
This episode analyzes the geopolitical fallout from the US ban on Anthropic’s Claude models, highlighting G7 discussions on AI sovereignty and access. It covers Noam Shazeer’s move from Google to OpenAI and the rise of alternatives like China’s GLM-5.2 and Cursor’s Composer. The discussion emphasizes how enterprises are adopting multi-model routing and open-source solutions to mitigate costs and dependency on restricted frontier models.
Topics discussed
Intro, announcements, and show notes
G7 AI meeting: US leadership and global cooperation
Reactions to G7, Mythos ban, and SK Telecom
Noam Shazeer leaves Google for OpenAI
OpenAI sunsets Pulse feature
Sponsors: KPMG, Section AI, AssemblyAI, OutSystems
Post-Fable ban landscape and open-source shift
New Chinese models: Kimi K2.7 and VibeThinker
GLM 5.2 release and benchmark performance
Microsoft fine-tunes DeepSeek for Copilot
Cursor Composer 2.5 vs GPT-5.5
OpenRouter Fusion API and model routing
Harvey AI, cost strategies, and conclusion
Listen ad-free on Castria