MIRI researcher Harlan Stewart challenges OpenAI's recent safety and alignment webpage, arguing that it relies on confusing timelines and unproven assumptions. Stewart urges the AI developer to take greater responsibility for proving its superintelligence safety strategy.