deep-research

Multi-source research coordinates Web search, code analysis, and cloud model consultation to produce a validated research report with synthesized evidence and an architecture decision record.

Validating external research against a codebase involves cross-referencing multiple web research reports with actual code constraints, categorizing findings as directly usable, needing verification, or project-specific exceptions.

Synthesizing findings from multiple deep research tools requires parallel execution of identical prompts, followed by expert review to identify logical flaws, weak evidence, and disagreements before generating a final synthesis document.

Multi-source synthesis is needed when technical questions require diverse evidence, design decisions need validation, or external cloud model consultation is necessary, rather than for simple searches or confirming existing conclusions.

Consulting cloud models involves summarizing the local background, sending it to the cloud model, backfilling the returned results, and finally synthesizing the combined insights into a comprehensive research document.

Limitations include dependency on multiple external deep research services, manual export of third-party reports, and the necessity of codebase validation to fill implementation gaps rather than trusting external outputs directly.