The point is that LLMs are doing human-ish things on their own, and that's what the value in them is, so OP is bringing attention to the fact that the tradeoff is "thinks dynamically, sometimes makes mistakes". Computers have never been built to think dynamically before.
"think dynamically" is the type of discourse TFA denounces, though. The thing designated by "think" is already largely unclear when you think about it, "think dynamically" sounds like self-improvement-coach nonsense.
