deep-research

Systematic research maps core questions to required evidence types, applying a defined plan across sub-questions. It organizes sources, notes, and final reports while maintaining disk-based memory for a reproducible workflow.

Keeping context lean requires using the filesystem as working memory. Scraped content, extracted notes, and source metadata are saved to disk rather than held in context, making research resumable and searchable.

To structure research, create a topic directory and draft a plan with sub-questions and a sourcing strategy. This plan defines the breadth and depth required to gather empirical data and primary sources.

Yes, disk-backed workflows support resumable investigations by saving files like plan, sources-index, notes, and report directly to disk. This ensures research remains searchable and reusable across sessions.

Generating a citation-backed report involves organizing gathered primary sources, methodologies, and data points according to the initial plan. The process returns a structured final document with transparent sourcing.

Avoid plan-driven research for simple, single-query lookups that require no empirical data mapping or multi-source investigations. The systematic overhead is unnecessary without the need for traceable notes and reproducible workflows.