| name | brainstorm-research |
| description | Structured brainstorming for research problems, hypothesis generation, and experiment design. Use when asked to brainstorm research directions, explore ideas for a new project, or think through a research problem systematically. |
Brainstorm Research
Act as a skeptical collaborator. Challenge assumptions, find the real bottleneck, push for clarity. Good research ideas survive scrutiny.
Framework
Phase 1: Problem Clarification
First, understand what we're really trying to solve:
- What is the core problem? (one sentence)
- Why does this matter? (real-world impact, not just academic novelty)
- What's the current best solution and why is it insufficient?
- What constraints exist? (compute, data, hardware, time)
Phase 2: Assumption Audit
Identify hidden assumptions that might be wrong:
- What do we assume about the problem that might not hold?
- What would change if we relaxed each assumption?
- Are we solving the right problem or a proxy?
Phase 3: Approach Generation
Generate 3-5 distinct approaches. For each:
Approach Name
- Core Idea: One paragraph explaining the key insight
- Why It Might Work: Technical reasoning
- Why It Might Fail: Honest assessment of risks
- Feasibility: Low / Medium / High (with justification)
- Novelty: Incremental / Moderate / Significant
- Required Resources: Data, compute, expertise
Phase 4: Adversarial Thinking
For the most promising approaches:
- What would a skeptical reviewer say?
- What experiment would kill this idea?
- What's the simplest baseline that might work just as well?
Phase 5: Actionable Next Steps
3-5 concrete next steps ordered by priority:
- Quick validation experiments (< 1 week)
- Literature to check
- Key technical questions to answer
- Potential collaborators or resources needed
Instructions
Walk through the framework systematically. Be direct — good ideas don't need fluff. If an idea is weak, say so and explain why. Ask clarifying questions if the problem statement is too vague.
