>I'm a data scientist, and google asked me to sum all values of nodes at each height of a tree. I had to implement the tree, bfs, and the algo (which was easy once you have bfs) in a 25 minutes, minus any talky time.
THIS is the problem. The (unreasonably) timed nature of the exercise means that the only people that will do well are the people that prep heavily for this specific skill, in much the same way that people about to take the LSAT are likely to do better than it than actual fucking lawyers with proven experience.
Now take this analogy one step further: why aren't lawyers asked stupid timed LSAT questions at every job interview? Because they take a really challenging professional exam called the bar, which is a strong base level guarantee of actual knowledge and competence. Software Eng. have been fighting this kind of credentialing because somehow tech is "top innovative" for such standards. Hence the status quo, and why companies like TripleByte see opprtunity.
> Now take this analogy one step further: why aren't lawyers asked stupid timed LSAT questions at every job interview?
Because which law school they went to is on their resumé and the average LSAT scores for each school are trivially available, tracking very, very closely with school prestige.
Time (and being watched) do make it very unrealistic, but the other thing missing is that programming now is not like in the 1970's, you don't need to memorize much of anything. If you forget something, you look it up, it takes 30 seconds to remind yourself, "oh, yeah, the queue", and then you go. Testing if you can do it without looking it up is not testing the skills you actually need in order to code well.
THIS is the problem. The (unreasonably) timed nature of the exercise means that the only people that will do well are the people that prep heavily for this specific skill, in much the same way that people about to take the LSAT are likely to do better than it than actual fucking lawyers with proven experience.