Berges Brainstorm

Berges Brainstorm

Divergent ideas, not safe ones. Pick Brainstorm from the sidebar when you want range, not a single best guess.

Brainstorm is the interceptor for when you want options. Pick it from the sidebar and the deep-thinking layer switches into a mode tuned to push for distinct, varied ideas instead of converging on the most plausible-sounding answer.

Most AI assistants quietly average toward the median. Brainstorm pushes the other direction: more spread, more variety, more rough edges. Useful precisely when the median answer is the wrong answer.

How it works

What changes when you pick it.

Higher randomness

Brainstorm uses a high sampling temperature, which gives the model more freedom to take less-likely paths. That's what produces variety instead of a polished single answer.

A divergence-first system prompt

On top of the model, Brainstorm layers a system prompt that asks for multiple distinct angles, with no early convergence. The framework is the proprietary part.

Thinking-tier only

Brainstorm only kicks in on escalated turns. The fast layer stays out of it, so easy questions keep their snappy default behavior.

Reach for it when

Brainstorm is the right pick.

  • Naming products, projects, or features
  • Generating angles for an article, deck, or talk
  • Exploring options before committing to one
  • Coming up with counterexamples or edge cases
  • Any moment where "give me five different ideas" is the right ask
Try it

Example prompts.

Put it to work

Things to do with Brainstorm.

Questions

Things people ask about Brainstorm.

When should I not use Brainstorm?

When you want a single, confident answer. Asking a Brainstorm-mode model "what is 17% of 240" will get you a correct answer with two unnecessary alternatives, which is not what you needed.

Are Brainstorm's ideas always good?

No. Higher randomness produces variety, and variety includes some misses. That's the point: you pick from a wider pool. If every idea were polished, the pool would have collapsed back to the median.

Can I refine an idea after brainstorming?

Yes. Switch back to no interceptor (or Reasoning) and ask the model to develop one of the ideas in depth. The conversation history carries over.