I did this recently with my unsorted bookmarks! It was the first time I used parallel API calls. Ten gpt-4-nano threads classifying batches of ten bookmarks ripped through 10,000 bookmarks in a few minutes.