Brainstorming and ideation

The task

You need options, names, angles, product ideas, campaign hooks, or alternative solutions. The model is not deciding yet. It is expanding the search space so a human can choose.

Why the old approach is outdated

“Give me 20 creative ideas” usually produces a list that sounds plausible and samey. The model has no reason to know the market, constraints, taste, or unacceptable directions.

Better brainstorming prompts provide the walls of the room. Constraints create better variety because the model can explore tradeoffs instead of filling space.

If the first output misses the mark, use prompt iteration to change one constraint at a time instead of asking for a total rewrite.

The current approach

Generate 12 product-content ideas for a prompting site.

Audience: working product, engineering, and operations people who use AI at work.
Constraints: no model-news posts, no thin template libraries, no claims that prompting
solves judgment problems by itself.

For each idea, include the reader problem, why it is durable, and one example section.
Group similar ideas instead of padding the list.

When to use a different model

Use Claude when the ideas need a quieter editorial tone or structured synthesis from a source pack. Use Gemini when the brainstorm depends on images, video, or other multimodal inputs. Use Mistral when you want to compare against a compact model family already present in your product stack.

What to avoid

• Asking for a fixed number when quality matters more than count.
• Using “be creative” without explaining what boring looks like.
• Letting the model pick the strategy and the ideas in the same pass.
• Keeping duplicates because they arrived with different wording.