It's funny because LLMs are actually horrible at logic.
Current AI is actually good at search, but I can see how people mistake one for the other in the context of math, for example.