Brainstorm Ideas

# Brainstorm AI Safety Research Ideas

Interactive collaborative brainstorming of AI Safety research directions, tailored to team constraints and participant profiles.

## Setup

Load project configuration:

```bash
uv run python -m saim.config.cli show-scoring

Load team profile:

uv run python -m saim.config.cli show

Load participant profile (if available):

uv run python -m saim.config.cli show-participant

If the participant output shows "NO_PARTICIPANT", guide the user conversationally:

No participant profile found. To tailor ideas to your constraints, tell me about:

1. Your background and experience level
2. Technical skills (programming languages, ML experience)
3. Available compute resources
4. Time budget for the project
5. Expected deliverables

Or type "skip" to brainstorm without constraints.

Save these constraints for use throughout the session.

Load previous ideas for divergence (avoid duplicates):

uv run python -m saim.pipeline.memory

Echo the active context:

Brainstorming with: team=[team_type], participant=[name or "none"], previous_ideas=[count]

Interactive Brainstorming Loop

Enter collaborative mode. For each user input:

If the user provides a topic, area, or problem direction (FR44, FR45):

Generate 3-5 research idea sketches tailored to:

• The specified topic/area/problem
• Team constraints (compute, skills, time from team and participant profiles)
• Previous ideas (avoid duplicating titles from pipeline memory)
• KB context (reference relevant papers/findings if KB is available)

For each idea, provide:

• Title: Concise, descriptive
• Problem: What gap or question this addresses
• Approach: High-level methodology
• Why it fits: How it matches the participant's constraints
• Estimated scope: Rough time/compute requirements

If the user asks a research question (FR48):

Assess whether the question has been addressed in the literature:

1. Search for relevant work using citation tools:

uv run python -m saim.verification.citation search-crossref '<query>'

uv run python -m saim.verification.citation search-s2 '<query>'

2. Summarize findings: what's been done, what gaps remain, and whether the question is open or largely resolved.

If the user wants to refine or iterate (FR46):

• Accept feedback on proposed ideas
• Strengthen weak aspects
• Combine elements from different ideas
• Adjust scope to better fit constraints
• Generate variations on promising directions

If the user wants to save an idea:

Write the idea as a markdown file to data/ideas/ following the existing naming pattern. This makes it available to future pipeline runs via pipeline memory.

Continue the loop until the user indicates they're done (e.g., "done", "exit", "that's enough").

Session Summary

When the session ends, provide:

• Count of ideas explored
• List of any ideas saved to data/ideas/
• Suggestions for next steps (e.g., run /evaluate-idea on favorites, run full pipeline)