The discussion revolves around leveraging tools like calculators and coding for calculations instead of relying solely on human memory. Comments suggest that while humans juggle a limited number of tasks in short-term memory, this memory could also serve as a tool. The balance between the capabilities of LLMs (Large Language Models) and human cognitive limits is emphasized, proposing that smaller models might be sufficient if they focus on tool usage and basic language understanding.