Tag
An analysis piece discussing why persona depth in LLM systems fails to improve output diversity, and how isolating readers with broad personas yields better convergence evidence.