Working with tables of data feels more right, more like programming then the usual shell. I also work quite often with CSV and it's nice to be able to manipulate/filter it like any other data that nu is able to read (just: open file.csv | find "foobar" | select name message).

Not quite a straightforward as nu seams to be, but I used to have ipython as my shell when I did a lot of work with CSV. It’s “not a shell” but includes some shell like functions to make listing dirs and such easier.