But that's really what I meant. When you say the limitation on processing is not in the math. I would say it is a mathematical limitation of processing because they had to choose a math that works on parts of words instead of letters due to the limitation of the power of the math that can be done for training and inference.

They chose to use some limiting math which prevents the LLM from being able to easily answer questions like this.

It's not a limitation of math in general. It's a limitation of the math they chose to build the LLM on which is what was going through my head when I was writing it.