But is the information really scanned and preserved for direct access? Or is it just trained into an LLM so that we can only get fuzzy answers about the information?