An AI pretraining researcher resigned from Anthropic today, and his exit thread is interesting. His claim: the people building frontier AI genuinely believe it could kill us all by 2030 and they're r...