Gurkenglas comments on [AN #95]: A framework for thinking about how to make AI go well