We're excited for foundational models to improve because we hope it will unlock a lot more use cases. Things like analysis after extraction, able to accurately extract extremely complex documents, etc!