Today's edition highlights a massive divergence in the AI landscape, led by the revelation that Chinese military researchers are using model distillation to bypass US export controls. While Amazon accelerates its infrastructure buildout with a colossal $220 billion capex target, the market faces its first major systemic casualty as Leopold Aschenbrenner's highly leveraged AI fund is liquidated. Meanwhile, rigorous new research challenges industry dogmas, proving that 1990s-era keyword matching beats modern agentic retrieval at scale, and that model self-reflection loops actually degrade performance compared to simple repeated sampling.