AI Outperforms Doctors in the ER, Anthropic Left Off Pentagon's List Again, AI Bans Spread

Today we're covering the biggest AI stories of May 4th, 2026. A Harvard study published in Science put a 2024-era OpenAI model through 76 real ER cases and it outdiagnosed two attending physicians at every stage — including flagging a rare flesh-eating infection in a transplant patient 12 to 24 hours before the treating doctor caught it, with reviewers unable to tell which diagnoses came from the AI and which came from humans. The Pentagon formalized its AI partnerships with SpaceX, OpenAI, Google, Nvidia, Microsoft, AWS, and Oracle — notably excluding Anthropic despite the White House privately wanting priority access to Mythos, with the DoD calling Anthropic's blacklist a "separate national security moment." Plus, Maryland became the first U.S. state to ban AI-driven grocery pricing, fining stores up to $25,000 for using personalized shopper data to mark up prices — and today's community workflow comes from a new real estate investor in Finland who used Gemini, Claude, and Codex together to build a custom property market analysis tool from scratch after finding zero transparency in his local market.