Hacker News new | ask | show | jobs
by an_opabinia 1798 days ago
In my experience, for every engineering subordinate that turned out to be good that I didn't bother testing with leetcode, there were 19 who sucked and also failed their coding test, or sucked and we should have tested them. So far, among people who studied programming as their major in college, the best predictor was the prestige of the university they attended.
2 comments
mattmanser 1798 days ago
If you've hired over 20 people, and 19 of them have been bad, it just sounds like there's something wrong with your process. It also sounds like you're hiring a lot of novices, which isn't what's being discussed really.

The uni thing is not exactly a revelation. It's the same for every single field/job. Better uni, usually better worker.

All you're actually saying is that novice programmers straight from uni mildly correlate in ability with SAT scores.

Nothing surprising in that.

link
bryanrasmussen 1798 days ago
>If you've hired over 20 people, and 19 of them have been bad, it just sounds like there's something wrong with your process.

I think this is part of the reason for the widespread usage of take home / coding assignments etc.

If you have a bad process making people do some work improves your result.

I remember the first company I had with a friend in the late 90s and our process sucked. It was embarrassing, although thinking about it we still had a 50% success rate in technical quality of people we hired. If we had given tests to the people we didn't have someone to vouch for it would have meant we did not make the mistakes we did. And by saying we had a bad process we had a bad firing process as well. The bad hires we made really were catastrophic because we couldn't handle any part of the process.

link
TheCoelacanth 1798 days ago
Yes, the thing that is wrong with the process is that they didn't check to see if the person was able to write code.
link
BlissWaves 1798 days ago
I like leetcode as a hiring metric and I think it's the best measure to test out an engineer. Unless a company uses it basically to find people that implement the fastest algorithm(read: have memorized it already) in 30 minutes it can be a very very effective way of screening.
link