What does an AI tl;dr summary afford that the writer's authored abstract doesn't (especially in cases where the writer uses the bot to write their abstract in the first place) other than satisfying an idle curiosity in how LLMs collocate the same input differently?