3 ms·

(Talking about several-hour long take-homes, not FizzBuzz problems). It is cruel, but it is numbers game. The take-home is easier for hiring managers given the

by liuliu 4y ago

(Talking about several-hour long take-homes, not FizzBuzz problems). It is cruel, but it is numbers game. The take-home is easier for hiring managers given the macro environment and would help to cull out false positives (hence "numbers game"). Our industry has way passed the point of not playing these numbers game. The most friendly ones I've seen is a "take-home" done onsite with generous time given, and an engineer can answer questions. It helps to simulate work environment as much as possible. The upfront cost is extreme for the companies though: you put up your engineers time for these onsites. You did end up with good hires, but the time limit is real (v.s. take-home), and that shoot up your false negatives through the roof. But it does respect candidate's time, and friendly. For companies can afford (in the sense that good candidates just walk away), "take-home" portion also helps "alignment". If you spend extra time on it and really want to get in, you are probably ideal (or desperate, but still ideal, gives the company more leverage). These are given no cheating is happening though. I have to admit that I never did good on these "take-home" tests. Setting up environment normally was not an issue, just without an engineer to clarify constraints, I often end up with less-than-commonly-expected solutions and make the interviewer hard to judge the end-result (most "take-home" from interviewers' perspective are very structured like exams, you get x points here, y points there, if you go off the expected path, it was hard to score these points). ChatGPT probably will do a better job than me at this these days though.

sokoloff 4y ago
One of my most enjoyable past interviewing experiences was doing a take-home for some game company (probably Accolade). They shipped it as a zip file (late 90s), and there was a clear "todo" function. The task was to navigate an airplane across the map from a starting square to a destination square while avoiding mountain squares. The airplane always moved ahead 1 square each move and you could tell it to turn left 45°, stay straight, or turn right 45°. If you compiled it without writing the function, the airplane just flew straight. Perfectly reasonable task, IMO; it's even relevant to the gaming domain; you could do a simple version in an hour; I spent about 2 hours as I wrote a test harness and let it run overnight keeping score of how many maps it succeeded on and how many it failed on. The interviewer used my submission in the on-site as well to ask a few questions about it. Your numbers game comment is spot-on. From the employers' end, the hiring market problem looks like: There's high pay for software engineering roles. There's no effective credential mechanism. (I think this good, but it's not helping employers here.) Because of the combination of the first two, lots of people are motivated to try to land a job. Because of that, lots of people who aren't really qualified enter the pipeline. Because they're not really qualified, they keep applying over and over, while the qualified eventually get hired. Because of that, the applicant pool has a massive adverse selection problem (the average applicant is *much worse* than the median employee) and companies have to find a way to filter out the worst-of-the-worst before applying significant manual effort.

by liuliu 4y ago

sokoloff 4y ago

One of my most enjoyable past interviewing experiences was doing a take-home for some game company (probably Accolade). They shipped it as a zip file (late 90s), and there was a clear "todo" function. The task was to navigate an airplane across the map from a starting square to a destination square while avoiding mountain squares. The airplane always moved ahead 1 square each move and you could tell it to turn left 45°, stay straight, or turn right 45°. If you compiled it without writing the function, the airplane just flew straight. Perfectly reasonable task, IMO; it's even relevant to the gaming domain; you could do a simple version in an hour; I spent about 2 hours as I wrote a test harness and let it run overnight keeping score of how many maps it succeeded on and how many it failed on. The interviewer used my submission in the on-site as well to ask a few questions about it. Your numbers game comment is spot-on. From the employers' end, the hiring market problem looks like: There's high pay for software engineering roles. There's no effective credential mechanism. (I think this good, but it's not helping employers here.) Because of the combination of the first two, lots of people are motivated to try to land a job. Because of that, lots of people who aren't really qualified enter the pipeline. Because they're not really qualified, they keep applying over and over, while the qualified eventually get hired. Because of that, the applicant pool has a massive adverse selection problem (the average applicant is *much worse* than the median employee) and companies have to find a way to filter out the worst-of-the-worst before applying significant manual effort.