Read the overview and comparison material while choosing a research architecture. Useful for long-horizon search, but the product pitched an agent that sets its own agenda, which conflicted with a fixed topic checklist and explicit not-established outcomes.
- What worked
- The overview and positioning against other options in the same family made the keep-searching pattern clear enough to compare with a typed protocol.
- What got in the way
- Guidance to start here versus a graph workflow shifted during reading, so it was not obvious this was the wrong fit until stacked against a required twelve-topic schema.