be-creative

What problem does it solve?

Alignment-trained models tend to produce the safe, obvious “centroid” answer, so brainstorming often yields generic ideas that feel like small variations of the first response.

Core Features & Use Cases

• Verbalized Sampling (VS): Generates multiple candidates with explicit probability estimates, then selects a winner rather than choosing the centroid.
• Tail-driven creativity: Forces exploration of low-probability but potentially high-fit ideas (e.g., hooks, names, metaphors).
• Rationale-first selection: Provides a short explanation of why the chosen option beats the alternatives.
• Best-fit scenarios: Product naming, taglines, story premises, headline/hook writing, and divergent ideation under constraints.

Quick Start

Ask for five diverse options with brief probability-aware selection rationale, then tell the model to “apply verbalized sampling and pick the best non-obvious winner.”