Slower paths means loops here for humans where the output of a neuron gets feed back into itself. The fastest path = a feed forward neural network without loops.

LLM’s operate strictly feed forward neural networks.