Comment by bulder

Seems silly to assume chain of thought analysis is the only way to monitor a model's actions instead of, monitoring the actions. What about when "thinking" is turned off, like what OpenAI lean to by default for free users?