I ran my non-neural assistant against small LLMs and it lost badly. Here's the honest table.

CAEL is a chat assistant I've been building whose core is an integer cellular automaton plus a set of symbolic engines — maths, a Python checker, memory. There is no neural network anywhere: no weight