“Ben Thompson of Stratechery asserts that Anthropic's confrontational decisions are justified internally by a sincere belief that they are the only ones sufficiently concerned about AI safety.”