The post reflects on when memory was primarily measured in kilobytes, discussing how that limitation necessitated the development of highly efficient algorithms in computer vision. With AI and modern computing power, there is a sentiment that while these old algorithms may seem obsolete, their core principles could still be relevant to today's advancements, particularly in the realm of Large Language Models (LLMs). There’s also curiosity about how these old techniques would perform if applied to contemporary hardware and software environments, hinting at a potential revival or re-evaluation of past innovations.