Research an Answer

1. Frame the decision or question, required freshness, relevant versions, and what evidence would change the conclusion.
2. Inspect user-provided and repository-local artifacts first. Treat canonical project documents and raw results as the source of truth for local claims.
3. Search external sources when needed. Prefer current official documentation, standards, source code, papers, and first-party data; use secondary sources mainly for discovery or competing interpretations.
4. Record publication dates, relevant software versions, methodology, and source limitations. Verify drift-prone claims against current primary sources.
5. Reconcile conflicts by comparing definitions, scope, samples, dates, and methods rather than choosing the most convenient source.
6. Adversarially challenge the emerging conclusion: search for credible contradictory evidence, alternative definitions, negative results, confounders, and boundary conditions. Apply comparable scrutiny to sources that support and oppose it.
7. Synthesize in plain language. Separate sourced facts, analysis or inference, assumptions, and recommendations; cite claims close to their evidence.
8. For statistics or experiments, explain effect size and practical meaning, distinguish main effects from interactions, avoid equating non-significance with no effect, and state what remains inconclusive.
9. End with the decision-relevant conclusion, strongest supporting and opposing evidence, limitations, and concrete next steps when appropriate.

Contract