Sex, AI, and the Apocalypse

A high-profile resignation from Anthropic by researcher Jacob Coxon, who warned that his colleagues "earnestly believe that [AI] could kill us all by the end of the decade," thrust apocalyptic AI rhetoric into the mainstream. That claim was publicly echoed by Anthropic leadership and amplified by politicians and tech moguls. Those warnings trace back not to generic technologists but to a distinct twenty-five-year-old "rationalist" subculture centered on LessWrong, the Machine Intelligence Research Institute (MIRI), and personalities like Eliezer Yudkowsky. That community grew around a doctrine of clearer thinking, a secular liturgy (Secular Solstice) organized by Raymond Arnold, and canonical writings (the Sequences) that recruited technologists into a worldview prioritizing existential risk mitigation. Its demographics skew heavily male, autistic or neurodivergent, and technically trained, and its founding narratives and leaders - Yudkowsky’s early transhumanism and MIRI’s long-standing concern about misaligned superintelligence - have shaped both belief and recruitment into contemporary AI safety institutions.

The social ecosystem producing these warnings is deeply entangled with troubling specifics: a Harry Potter fanfic recruitment pipeline, psychological workshops described as coercive, programs advocating eugenic child-enhancement, sex-and-culture controversies, a sect that retreated to boats and faced links to multiple deaths, and bloggers who promoted anti-democratic ideas that reached high government offices. These cultural features - salvation stories, authoritarian sympathies, fraught sexual dynamics, and experimental social practices - are embedded in organizations now asking for public trust on AI governance. Assessing the credibility and policy weight of doomsday claims therefore requires scrutiny of the community’s intellectual lineage, social practices, and political influence.