This, in the end, was what motivated me to reintroduce “Critique Claude”/”Guide Gemini”/”Oversight o3″.[3] That is, a secondary model call that occurs on context summary whose job it is to provide hints if the model seems stuck, and which is given a system prompt specifically for this purpose. It can be told to look for telltale signs of common fail-states and attempt to address then, and can even be given “meta” prompting about how to direct the other model.
Funnily enough this reminded me of pair programming.