Research OS — Evidence-Driven AI Research Workflow

The problem

AI makes research dramatically faster, but speed creates new problems: duplicated work, unclear authority, unsupported conclusions, model drift, expensive context reconstruction, and findings that become detached from their evidence.

What I built

Research OS is a structured AI-assisted research workflow designed to move from question to evidence to decision while preserving state and provenance.

The system combines:

exploratory research

model routing

structured evidence capture

machine-readable state

staged escalation

independent review where useful

durable research artifacts

explicit closeout and recovery

Local models are used where they are sufficient, with more capable systems deliberately introduced when the task justifies them.

Result

The workflow has been applied across software architecture, technical research, security research, market analysis, and other long-running investigations.

Its purpose is not to automate judgment away. It is to make research easier to resume, inspect, challenge, and reuse.