OpenAI's AI autonomously disproved a geometry conjecture that stumped mathematicians since 1946 — using techniques no one in the field had thought to apply. Then: an Anthropic co-founder predicted an intelligence explosion by 2028. Meta was caught harvesting employee data to train AI before firing 8,000 people. Trump killed the AI safety order after three CEO phone calls. And GitHub got hacked in 18 minutes through a developer extension.