research

Parallel scientist orchestration works by decomposing complex research goals into 3-7 independent stages, each with explicit focus, hypotheses, and scope. Tailored scientist agents execute stages concurrently while verification loops cross-validate findings before final synthesis.

Manage research sessions using built-in commands to resume, list, report, and cancel investigations stored under .omd/research directories. Session state tracks parallel scientist progress, verification loops, and synthesized reports across multiple research goals.

Run fully autonomous research loops using AUTO mode with configurable iteration control. Model routing tiers automatically assign low, medium, and high complexity work to appropriate scientist agents while enforcing concurrent scientist limits and verification loops.

The best way to decompose research goals is breaking them into 3-7 independent stages with explicit focus, hypothesis, scope, and tier definitions. This structured decomposition enables parallel scientist dispatch, verification loops, and comprehensive evidence-based synthesis.

Parallel scientist orchestration applies to comprehensive domain investigations requiring structured stages, session tracking, and AUTO mode orchestration. The decomposition, execution, verification, and synthesis framework generalizes to any complex research domain beyond codebase analysis.

Avoid parallel scientist orchestration for simple research tasks that do not require structured decomposition, verification loops, or session state tracking. The overhead of coordinating multiple agents and maintaining .omd/research directories suits complex, multi-stage investigations only.