Agree on the reproducibility aspect, but not on the static medium of PDF

The big issue with notebooks is reproducability; I often have difficulty getting a notebook running that we written by a colleague at the same organisation a few months ago.

Given their expanded capabilities, notebooks are inherently going to be harder to reproduce in their orignal fidelity than a PDF, which is fully described by a self-contained specification, though how one is meant to read a specification distributed as a PDF leads to other questions. Being open makes reproduction possible, but it is still a significant hurdle.

I think static documents such as PDF still has a role to play, even if it is just a medium to determine one's interest in some work before investigating the code.

The web is an interesting publishing medium, forming a sort of middle ground between static PDFs and the fully dynamic notebook. I see more journals embracing it as a distribution mechanism; some even allow for interactive plots powered by JavaScript. So perhaps that will be the future direction.