Comment by crystal_revenge

> A lot of commenters causally speak of "false negatives" as if they were random

They also don't talk about, what my experience has shown me, is an extremely high false positive rate.

I've passed these style interview for a few large companies and those places easily had the least skilled developers I've worked with. Similarly, I've been shocked by the lack of technical skill in many ex-FAANG coworkers I've had. Their are some exceptions, but those are typically people who worked for a FAANG nearly a decade ago.

In the early days when Google was doing more CS style algorithmic thinking tests, it might have been a better measure, but this world of "grinding leetcode" as a skill provides very poor signal on developer skill.

For an MLE position Meta currently requires two leetcode mediums to be completed within 40 minutes and the solution must be absolutely optimal, and this is as a screen. This can only be reasonably accomplished by studying all 100+ of their problems on leetcode so you know the answer before hand. I do think thoughtful algorithm design interviews can be high signal, but in it's current form you can't test anything other than "how many hours did you study this?"

Most of the smartest people I've known and worked with have much more interesting things to do with their time than grind leetcode. Additionally, at this level of grilling, you're not even thinking anymore. In fact, the interviewers seems genuinely surprised if you give a correct answer that is not exactly performed in the canned way expected. White boarding algorithm interviews can be great, but what we have to take is a sad facsimile of what was originally being tested.