[jojobas]

The point is never to have a working graph clusterization algo. It's to see that you not only theoretically know the difference between deque and heapq, but also know which is the right thing for a particular task you're working on.

I've had people claiming 15 years of dev experience fail to find the 20th last comma in a file and not even being close after an hour.

You only start getting it once you see how bad 9 out of 10 people really are.

[batshit_beaver]

Is it that they're bad, or is it that your tests stumble upon a small area in the huge realm of software engineering that those particular engineers struggled with on those particular days? Or, if you're asking these people the same group of questions and seeing a 90% failure rate, maybe your chosen questions have little overlap with common engineering tasks?

Why is it so hard to believe that even after 15 years of producing useful, quality software someone might still struggle with a random problem in an interview setting? What level of arrogance leads to this holier than thou thinking?