“What’s fallacious with LLMs and what we must be constructing as a substitute”