becomes a pain when dealing with docs in bulk, or with docs that have multiple pages. Token length limits with GPT with a single request and I see a loss of context accuracy when extracting insights in multiple requests.