AI researcher whose apocalypse warning went viral outlines his disagreements with Anthropic leadership

Jacob Coxon, the researcher who left Anthropic warning AI could kill us all by the end of the decade, did an X AMA today and doubled down hard. His whole argument is that Anthropic leadership justifies racing to the frontier with this "well someone's gonna do it anyway" logic, and Coxon basically says that's a cop out, his exact line was "if the race is inevitable you should not contribute." He also went after their push into recursive self-improvement, letting AI models improve themselves and build their own successors, calling that the genuinely reckless part.