In recent years, notebooks have become a mainstay of the data science toolkit. Although the literate programming paradigm, combining source code and documentation in a single file, may be well suited to modeling and analysis, notebook users have not yet reaped the full benefits of this paradigm for several reasons – including the hidden state. In this blog post, Barry McCardel and Glenn Takahashi, the founders of Hex, describe the challenges that arise when notebooks are backed by a kernel with a mutable state. They introduce a new reactive execution model for notebooks, which they implement and release with the latest version of Hex.