A Wharton School study suggests that when you ask AI for 100 ideas, they get more interesting the further down you go on the list. The technique involves a two-stage prompt: first, ask ChatGPT for "100 ideas for [topic]," numbered 1–100, with only a short title or name for each, including "the obvious, the bad, the weird, and half-baked fragments," where "quantity is the only goal." Second, ask ChatGPT to revise the list for maximum diversity, replacing similar ideas with ones from angles nothing else covers, aiming for "100 ideas where no two point at the same underlying concept." Optionally, a third stage can filter by quality: "What are the 10 most interesting ideas on the second list?" The author tested this for naming a coffee shop. The first 10 initial ideas included "The Daily Grind," "Bean There," "Brewed Awakening," "Central Perk," "The Coffee House," "Morning Cup," "Java Junction," "Common Grounds," "Cup & Bean," and "The Roasted Bean." After the second-stage prompt and optional third stage, the 10 most interesting names were: "Warm Noise," "Morning Object," "Public Living Room," "Moth & Match," "Localhost," "Borrowed Sugar," "Unfinished Sentence," "Blue Hour," "The Loading Bar," and "Sunday Weather."