Personally, when I write shell commands, they tend to be write-only code, because the shell isn't really suitable for anything more complex. So it's easier to think in code than it is to add the extra step of translating to English if it's something nobody's gonna see again anyway.
My counter to this would be - if I'm doing a thing, and I need to look back through my shell history a few days or weeks later to figure out how I did it (especially if something went wrong), seeing a more readable version is going to help me figure it out faster.
Granted, anything really nontrivial I'll usually just write a Python script for, but the point stands.