What's Wrong with Computational Notebooks? [pdf]

I wrote a lot of notebooks because my PhD supervisor liked using them as demos for our scientific software packages, which was for physics simulations.

But in my mind, most things that we wanted to show just weren't useful in notebooks, because they took too long to run. They're great for prototyping stuff like figure generation, and I used them a lot for that personally, and even for setting up the initial conditions for simulations they can work well, but they're terrible for doing a whole workflow of set up simulation -> run simulation -> process data in my opinion.