Re:Don’t do the long take home assignments like coinbase
by chipotle_coyote on Friday January 19, 2018 @12:23PM
(#16187928)
From reading the Reddit post, I think it's the combination of a take-home assignment plus the standard on-site whiteboard test battery that the OP was arguing against, and I admit I'm sympathetic to that. I remember interviewing at one place a few years back that had a relatively complex at-home assignment -- you had to write code that would hit various URLs that would return a string containing a long basic math formula (e.g., "1234-(3529(42-34)/2)+45", but going on for ~60 characters), write code that would parse that string to produce the result, then hit another URL pointed to by the result, while treating all the results as a node graph you needed to find the shortest path through. Granted, I'm one of those annoying self-taught programmers without a computer science background, so I probably found that a bit harder than someone recently out of school with these algorithms fresh in their mind -- but I got through the challenge correctly, wrote pretty nicely-structured and commented code to boot...which was, as far as I can tell, completely ignored* by the interviewers, who just asked me more on-the-spot algorithm questions instead.
I've been at other interviews in which the on-site interview is a walkthrough of the code I'd written, asking questions about why I made various choices. I think this is really a much better way to do programming interviews. It still weeds out people parroting code without understanding it, but it's a better example of what a candidate's actual code looks like -- and for the vast majority of programming jobs, that's probably more important than whether they've memorized algorithms that rarely come up outside other whiteboard tests.
I've been at other interviews in which the on-site interview is a walkthrough of the code I'd written, asking questions about why I made various choices. I think this is really a much better way to do programming interviews. It still weeds out people parroting code without understanding it, but it's a better example of what a candidate's actual code looks like -- and for the vast majority of programming jobs, that's probably more important than whether they've memorized algorithms that rarely come up outside other whiteboard tests.