Squeeze My Prompt
All tasks

I want to brainstorm ideas.

Every question is optional. The ones marked Recommended matter most.

For example: How can our 3-person bakery get more weekday-morning customers?

For example: Costs under $500, can start within 2 weeks, brings in commuters.

For example: We open at 6am; no delivery staff; near a train station.

For example: Loyalty card (didn't work); discounting.

For example: 12 ideas: half practical, half unusual.

For example: List, then rank the top 3 with one risk each.

For example: Commuters aged 25-45.

More detail, if you have it

How should the AI start?
Saved on this device only

How to brief an AI to brainstorm

Use this form when you need options: ways to win customers, names, gift ideas, ways around a problem. Brainstorming with AI has one known weakness. In a Wharton study published in Nature Human Behaviour, sets of ideas produced with ChatGPT’s help were significantly less varied in 37 of 45 comparisons across five experiments. The questions here are chosen to push against that.

  • Challenge. Phrase it as a question. Anthropic’s prompting guide and Google’s prompting guide both ask for clear, specific instructions.
  • What a good idea does. Your criteria let the AI judge its own ideas the way you will. Anthropic and OpenAI’s prompt guidance both advise stating what success looks like.
  • What’s fixed. Budget, hours, staff and place turn generic suggestions into ones you could act on. Google’s prompting guide advises giving the model the context it needs.
  • Already tried or ruled out. This stops the list filling up with what you have already done. Anthropic’s guidance for Claude Opus 5 recommends stating scope plainly.
  • How many, how different. Asking for several distinct ideas, with some structure, brings variety back: see a second Wharton study and research on verbalized sampling, which measured 1.6 to 2.1 times the diversity in creative writing when the model was asked for several responses at once, each with a probability.

There is deliberately no field for example ideas you like. Examples steer an AI strongly, so one could narrow the list. That is a reasoned guess, not a tested finding.

A worked example

A three-person bakery wants more weekday-morning customers. Their answers: the question, what a good idea must achieve, what is fixed, and what they have tried. The form composes:

I need ideas for this question: How can our 3-person bakery get more weekday-morning customers?

<context>
What's fixed: We open at 6am. No delivery staff. Next to a train station.
</context>

<task>
Brainstorm ideas that answer the question above. Make each idea meaningfully different from the others, not variations on one. Give each a short name and one sentence on why.
</task>

<constraints>
A good idea: Costs under $500, can start within 2 weeks, brings in commuters.
Already tried or ruled out, so suggest other things: Loyalty card (didn't work). Discounting.
</constraints>

If the information here is not enough to answer well, say so and tell me what you would need. If something missing would materially change the result, ask me up to 3 short questions first. Otherwise proceed and state your assumptions.

What the form adds for you

It asks for ideas that are meaningfully different from each other, each with a short name and a reason, because the research above found that asking for distinct ideas raises variety.