Makes it easier to explain why response style can vary.

Anthropic published research on how assistants may select a “persona” based on context. It explains that models aren’t hand-programmed like typical software but trained on large datasets to produce plausible behavior. This helps users better interpret style shifts in responses.