I think that's really the key thing, and I didn't realize it until only recently: org-mode fits a Jupyter mental model better than a Markdown (or outliner, etc.) mental model.
It's great for what it is, but what it is doesn't 1-to-1 map to much else.
I think this makes sense. If you are viewing it as marking up a static document, then a ton of the affordances in org-mode make little sense. Most of it is about interacting with a lot of data that happens to be represented in plain text.
For a lot of us, that focus on the plain text is important. As it helps keep us focused on all of the other affordances that we have grown accustomed to in emacs. As soon as you move things into a binary format, you are likely giving up on a lot of the other tools you have. There can be reasons to do this, of course, but for most data that you interact with at a personal level, most of those reasons have been overcome by the power of the personal computer.
It's great for what it is, but what it is doesn't 1-to-1 map to much else.