I've done quite a few "at home" programming challenges [1][2][3] as well as hacking challenges and most of my experiences have been overwhelmingly negative. The challenge is that "programming" is about 20% writing code and 80% other stuff (requirements, design, testing, security, documentation, meetings, beer, etc...). It seems like the proposed solution is not addressing the problem: Explaining how to solve a programming challenges you may have never heard of on the spot.
I do agree that interviews add a level of pressure, but what you're trying to gauge thought process. How does the candidate think about solving the problem; Can they iterate; Can we lower the space and/or time complexity; Can this be solved recursively, Are they receptive to input? A take home assignment does't help you figure any of that out. Usually, the team who is interviewing already knows the answer to the specific problem and all of the intricacies (based on having to solve it on an internal project or just because it's a question they've asked before), but the goal is to see someone thinking in real-time.
In my career, I've never had a time where I go off on my own, work on something and come back to the team with a solution. Even when I was writing TI-83 programs in High School to solve all of my science homework and finish 50 minute tests in 10 minutes; I still had classmates asking me about what I was doing along the way. There is always discussion and I feel like by sending me off on my own, all of that goes out of the window. You can't verify if I copied the code from anywhere else, you don't know if I had my buddy who is an engineer at another company break down the answer for me to spit it back to you, and you don't know if I paired with someone else to get the assignment complete. I'm obviously biased based on my experiences doing over 40 interviews in Silicon Valley/SF. My goal for this type of interview question would be to understand how the candidate thinks; I don't care if you write your answer in Brainfk[4].
I do agree that interviews add a level of pressure, but what you're trying to gauge thought process. How does the candidate think about solving the problem; Can they iterate; Can we lower the space and/or time complexity; Can this be solved recursively, Are they receptive to input? A take home assignment does't help you figure any of that out. Usually, the team who is interviewing already knows the answer to the specific problem and all of the intricacies (based on having to solve it on an internal project or just because it's a question they've asked before), but the goal is to see someone thinking in real-time.
In my career, I've never had a time where I go off on my own, work on something and come back to the team with a solution. Even when I was writing TI-83 programs in High School to solve all of my science homework and finish 50 minute tests in 10 minutes; I still had classmates asking me about what I was doing along the way. There is always discussion and I feel like by sending me off on my own, all of that goes out of the window. You can't verify if I copied the code from anywhere else, you don't know if I had my buddy who is an engineer at another company break down the answer for me to spit it back to you, and you don't know if I paired with someone else to get the assignment complete. I'm obviously biased based on my experiences doing over 40 interviews in Silicon Valley/SF. My goal for this type of interview question would be to understand how the candidate thinks; I don't care if you write your answer in Brainfk[4].
[1] - https://github.com/joeblau/sample-elevator-control-system
[2] - https://github.com/joeblau/sample-top-ten-tweets
[3] - https://github.com/joeblau/sample-url-shortner
[4] - https://en.wikipedia.org/wiki/Brainfuck