I think that the people not being impressed by cut and sort are approaching this from the Linux end of things, where those tools are nothing special at all. I guess we kind of expected that the data science wizards would be using fancier tools.
Yeah, well, people who have enough self regard to think of themselves as "wizards" are super unlikely to be able to actually do the day to day grind of getting, cleaning and preparing data for feature generation, which is about 95% of the job.
Another good weeder for a person claiming to be senior: discuss how you would fix the performance of the default R naive Bayes implementation in e1071. It's numerically more or less correct, but written by deranged ape-men who don't understand how computers work (a problem in a lot of the R ecosystem; in the Python ecosystem, the problem is nobody has yet written algorithms for X, which ends up being a very similar problem: aka it's your job to code up sane algorithms).
