colinlh/paperscientist/research-ideation

Research ideation

Review the user's profile, current project constraints and retrieved literature. If interests are broad, propose a small set of contrasting directions and one low-cost experiment for each. Do not fabricate a novelty claim.

For each candidate write a claim card:

• Observed limitation and supporting paper/section.
• Why this limitation matters for the target problem.
• Hypothesized cause and proposed intervention.
• What changes relative to the strongest existing approach.
• Conditions under which the idea should work, fail or become unnecessary.
• What measurement could refute it.
• Data, compute, implementation and time requirements.

Challenge the proposed mechanism with alternative explanations and nearest-neighbor prior work. When useful delegate an independent critique to a reviewer. Prefer a defensible small contribution over unsupported breadth. Keep candidate ideas distinct from the direction chosen by the user.

Use update_research_plan to preserve accepted hypotheses and evaluation questions. Do not turn planned experiments into reported findings.