True, though I have found that forcing (I use an agent skill to do this) an LLM's agent to document the reasoning behind each "decision" it makes seems to lead

True, though I have found that forcing (I use an agent skill to do this) an LLM's agent to document the reasoning behind each "decision" it makes seems to lead to better decision-making. Or at least, more justifiable decisions (even if the justification is bad).