What problem does it solve?
It turns open-ended technical curiosity into a structured, goal-driven research arc that iterates autonomously, verifies claims, and persists findings for later decisions.
Core Features & Use Cases
- Goal-driven research arcs: frames an arc, sets iteration and per-item goals, and keeps the thread coherent over multiple rounds.
- Parallel dispatch + verification: dispatches multiple research subagents when items are independent, then empirically verifies load-bearing claims before folding.
- Scrutiny and durable knowledge capture: runs adversarial scrutiny, then saves reports into refs/research/ and updates bead notes for long-term traceability.
- Layering without waiting: converts findings into new questions and automatically schedules the next research layer, supporting autonomous “keep going” loops.
Quick Start
Ask the agent to research a topic, include the question you want answered, and say you want it to stay autonomous while it iterates until the arc goal is satisfied.