Compset
A competitive-analysis tool where a model does the research and a guardrail keeps it inside limits. The log shows who decided each step: the model, the guardrail, or you.
In the recording
- You set the brief and the depth, then review the competitors it found before any research starts.
- The model researches each one. Every pass gets two web searches, and a third is blocked.
- When the model asks for more than the depth allows, the guardrail overrules it and the run carries on.
- The report ends with a record of what you changed and how often the guardrail stepped in.
Built as a Claude artifact. It needs Claude and a search connector to run, so this page shows a recording of it.