Why? What if LLMs could parallelize much of their reading and then summarize the findings into a markdown file, eliminating the need for complicated search?