One post argues that “P(doom)” bundles together two different probabilities. The author treats ruin as certain if artificial superintelligence (ASI) is built using anything resembling current techniques, or new techniques invented and managed by language models resembling Astra/Fable. But they express much less confidence about whether ASI will be built, citing uncertainty about future policy choices. They argue for a more hardline policy to prevent ASI and accuse accelerationists of overstating its inevitability.