I disagree pretty strongly. I wrote lots and lots of practical software as part of my undergrad. And the software I wrote was pretty diverse, including, file system drivers, image recognition, data mining / text classification, exploit utilities, etc. Most of the complex theory has been offloaded to graduate school to make way for "practical applications." Plus, I feel like my education set me up to be a little ahead of the job market, as data science was a track of my CS ten years ago, and you'd have left university with your own little scikit-learn library.