Reasoning in llms have catered long discussions...and since building word squares takes iterative reasoning and backtracking I thought let's ask llms to do it..The results were fascinating...Deepseek showcased profound reasoning whereas gemini failed but it's reasoning mimicked closely to humans...chatgpt tried but resorted to retrieval type answering where as claude hallucinated to the brink...showcasing how llms are not truly reasoning rather just presenting memorization in a way that feels like reasoning...