I don't think calling a model "relentlessly proactive" is positive anthropomorphism.

Spinning up 50 unnecessary subagents is exactly what I'd expect from a "relentlessly proactive" model.