#PyDataGlobal2023 just started and the very first talk was about good practices around Jupyter notebooks: write functions, use git, don't deploy them to production, etc.
We've been on this theme for *years*, and we keep insisting. Aren't we missing some key usability issues around the workflows we propose?
For example, functions: `%autoreload` is known to be flaky https://ipython.readthedocs.io/en/stable/config/extensions/autoreload.html#caveats and yet there's no good solution for developing library code and notebooks together.
