Why So Many AI Researchers Think the Machines Could Kill Everyone (wired.com)

A combination of rapid advances, recursive self-improvement, and agentic swarms are genuinely “spooking people” inside big labs.