In this video I cover three big moves in the AI world that all connect to the same theme: switching costs are collapsing, and the major labs are sprinting. First up is Anthropic’s new Claude memory import flow at claude.com/importmemory. It’s basically a two-step copy/paste that lets you pull your existing “who I am / how I work” context from ChatGPT, Gemini, or whatever you’re using, and drop it straight into Claude’s memory settings. The whole point is to make your first conversation feel like your 100th—because most people don’t leave a tool due to quality, they leave because onboarding is exhausting. Then I break down OpenAI’s GPT-5.4 release, and it’s not just a version bump. The headline is consolidation: instead of picking separate models for coding vs reasoning, 5.4 tries to do it all in one. OpenAI also claims big benchmark gains (83% vs 71% on an occupations eval), adds native computer use, introduces “tool search” for cleaner dev workflows, improves efficiency, and reduces factual errors compared to 5.2. Finally, I unpack the “Anthropic code red” situation: the Pentagon contract dispute, the supply chain risk designation, Trump publicly attacking Anthropic, and Dario Amodei’s leaked memo calling OpenAI’s Pentagon deal “safety theater.” It’s a messy, high-stakes story—especially with IPO rumors—where public sentiment and government relationships are pulling Anthropic in opposite directions.