Musings on AI safety, alignment, mech interp, AI governance and stopping robots from taking over the world (hopefully).