3 ms·

I understand there is some value in using a technical screen to see if a candidate has the viable expertise and skills required but how do we find that if every

I understand there is some value in using a technical screen to see if a candidate has the viable expertise and skills required but how do we find that if everyone is doubling down on generic competitive-programming style hiring sites? After crunching through many leetcode problems I see the problems they give have really poor APIs and many posted solutions people share have memory leaks and other problems that you would absolutely want to avoid on a production code base. Competitive programming is fun for sure but where is the data that it has a high correlation to job success? It seems to me like a completely different skill set. Writing fast, sloppy code in order to solve a puzzle is fun but it's quite another when your library brings down a production server because it leaks memory like a sieve.

nickysielicki 6y ago
> After crunching through many leetcode problems I see the problems they give have really poor APIs This. The real-world best solution to many leetcode problems is to implement an iterator over their broken interface so that <algorithm> can solve it. Because of how they constrain you, what you end up doing is just sloppily reimplementing portions of the standard library. That's something that you should never want to do in the real world. It completely selects for the wrong thing.
hysan 6y ago
leetcode is the StackOverflow for interviews. If all you do is copy paste the problem, you end up with a bad interview process. If instead you use the problem as a base, understand what signals it is trying to test for, and change it to implement your specific needs, you have a much better shot of gauging the right skills. Of course, that requires putting time into designing an interview pipeline full of things like rubrics, bias training, and sometimes realizing that leetcode questions don’t evaluate for the skills you need. Most companies aren’t willing to invest time into this and prefer to just copy <big co X> which is why you hear so many bad stories with what sounds like the exact same interviewers.

nickysielicki 6y ago

> After crunching through many leetcode problems I see the problems they give have really poor APIs This. The real-world best solution to many leetcode problems is to implement an iterator over their broken interface so that <algorithm> can solve it. Because of how they constrain you, what you end up doing is just sloppily reimplementing portions of the standard library. That's something that you should never want to do in the real world. It completely selects for the wrong thing.

hysan 6y ago

leetcode is the StackOverflow for interviews. If all you do is copy paste the problem, you end up with a bad interview process. If instead you use the problem as a base, understand what signals it is trying to test for, and change it to implement your specific needs, you have a much better shot of gauging the right skills. Of course, that requires putting time into designing an interview pipeline full of things like rubrics, bias training, and sometimes realizing that leetcode questions don’t evaluate for the skills you need. Most companies aren’t willing to invest time into this and prefer to just copy <big co X> which is why you hear so many bad stories with what sounds like the exact same interviewers.