Brainstorm 20 Ideas, Then Kill the Bad Ones
Get a wide, weird set of ideas — then a quick ranking so you can spot the 2-3 worth pursuing.
When to use this
When you need range, not safety — and you'd rather sift through 20 ideas than settle for the first 3 the AI offers.
The prompt
You are a creative collaborator with a wide range and zero attachment to your ideas.
- **What I'm trying to come up with**: [the brief — a name, a feature, a marketing angle, a gift, a side project]
- **Constraints** that must be respected: [budget, timeline, audience, any non-negotiables]
- **What's been done before** (so we don't retread): [...]
- **One adjacent territory** I'd like you to explore: [a vibe, a field, a reference]
Generate **20 ideas**, then rank them. Specifically:
1. **Round 1: Generate 20 ideas.** Number them. Spread the range — some safe, some weird, some uncomfortable. At least 3 should make me wince a little. Don't filter yet.
2. **Round 2: Rank them.** Sort into:
- 🏆 **The best 2** — what makes each strong, and the next step to validate
- 👍 **The interesting 5** — promising but need more thought
- 😐 **The middle 8** — fine, not great
- 🗑️ **The weakest 5** — and the one specific reason each is weak
3. **Round 3: One synthesis.** Combine two ideas into a 21st option that might be stronger than either alone.
No "this idea has merit" hedging. Be decisive.
What you'll get back
20 numbered ideas spanning a real range, sorted into tiers with reasoning, and one synthesized 21st idea combining the best of two.
How this is structured in English
Notice the English patterns this prompt uses — they're worth borrowing for your own requests.
- Make me wince a little Specifies a quality (slight discomfort) that signals real creative range. The 'wince test' filters out only-safe ideas.