research-assistant

An evidence-led research brief identifies the core decision, separates must-know from nice-to-know information, and structures inquiry into background, current state, and competing viewpoints. It gathers evidence using targeted web searches and fetches high-value sources directly to support executive decisions.

A structured research plan outlines inquiry topics including background, quantitative evidence, competing viewpoints, risks, and open questions. It parallelizes subtopics when appropriate, capturing each subtopic's fact, significance, and source to ensure comprehensive evidence synthesis.

Evidence synthesis uses web_search to locate likely sources, fetches high-value sources directly, and reads local files if provided. It captures each subtopic's fact, significance, and source, cross-checks claims, and presents clear caveats to produce a cohesive briefing.

Yes, you can include local files when gathering evidence for research analysis. The process reads local files if provided alongside fetching high-value web sources directly, allowing you to synthesize internal documents with external web_search results into a unified brief.

To ensure current information, the research plan includes at least one recency-focused query for the current year. This guarantees the evidence synthesis captures the latest current state and quantitative evidence alongside historical background data.

Automated evidence synthesis relies on locating likely sources via web_search and fetching them directly, which may miss paywalled or non-indexed primary sources. It cross-checks claims and presents clear caveats, but shallow or biased conclusions remain a risk if high-value sources are inaccessible.