You are building a research assistant that helps users explore a technical topic by finding relevant papers, reading them, and producing a synthesis with citations. The agent should browse paper sources, extract key claims, compare methods and results, and surface uncertainty when evidence is weak or conflicting. Users may ask broad questions like "summarize recent work on test-time compute" or narrow ones like "compare retrieval methods used in long-context QA papers."
Walk me through designing a research agent that browses, reads, and synthesizes papers.