Two AI agents reviewing each other's work sounds gimmicky until you try it.
I give both agents the same context and draft. Agent A works on it. Then I ask agent B to critique A's output and suggest what I should tell A.
What happens? They catch each other's mistakes. Sometimes they disagree and I have to mediate. Most of the time I just enjoy watching them work.

