Pentagon's Anthropic Ban vs OpenAI Deal & Google's New Nano Banana 2

Today's episode covers the Pentagon's blacklisting of Anthropic over AI safety disagreements, OpenAI's competing Pentagon deal with specific safeguards, Google's new Nano Banana 2 image model that tops leaderboards at half the cost, Alibaba's Qwen3.5 models matching Claude Sonnet 4.5 performance on consumer hardware, Perplexity's new 19-model orchestration system, and other significant AI developments including Karpathy's MicroGPT release and Google's robotics strategy shift.