The website is great and the examples (like getting distinct values of a table as a prerequisite investigation) really get the point across.

In my job I always end up with big notebooks of data exploration that get messy fast and are hard to share anything but the final result, having a canvas that embraces the non-linear nature is a great idea.

That has been my experience as well!

Aside from the non-linearity, what key features would make you use Kavla instead of a notebook?