So now we’ve reached the point where one AI needs to verify another’s step-by-step thoughts. Feels like the early days of code linters — only now it’s for reasoning chains. Honestly, not mad about it though… if LLMs are going to "think out loud," someone’s gotta fact-check the monologue.