It's not new - I did my first assignment around 18 years ago - and it's not necessarily a bad thing if done right. One problem is that assignments take too lon

It's not new - I did my first assignment around 18 years ago - and it's not necessarily a bad thing if done right. One problem is that assignments take too long and have over-optimistic estimates, which apart from taking up time can lead to feelings of self-doubt. Feedback on the assignments is lacking to no-existent, and this is an unforgivable sin in my book. Also, anything that looks more like a contribution to their own software than an academic test is a red flag. The other options noted in the article are also pretty bad. Whiteboards and fibonacci numbers, forget it. "Pair programming" in an interview is deceptive, nothing like normal pairing; it's a test (fair enough, but don't call it pairing). It's also problematic for me because I have a problem with my eyesight and any "accommodation" requested up front tends to be inadequate, leaving me at a disadvantage. For people with busy lives or family commitments, a short "interview prep homework" of 30-90 minutes would help the person to prepare, in their own time, in a relatively less pressured environment. The interviewer could then expand on this during the interview. I've done assignments like this where it proved a useful opportunity to brush up on a particular language (if cross-training or coming back from management), sometimes even enjoyed it. Also, it means recruiting is less of a numbers game - you might interview with a handful of companies instead of a dozen, choosing the ones you really care about. IMHO what we're seeing is a terrible implementation of a reasonable idea.