[stillsut]

This rejection is symbolic of a large problem emerging in the tech talent pool, similar to the "not-an-MBA" that we've seen ruin the management culture.

Large companies where you can learn a ton, and work with smart and motivated people, are filtering based on not entirely relevant CS trivia. They still have many viable candidates as anyone coming out of a university has four years where their job is explicitly to learn this stuff.

The stuff ends up being so esoteric, that rarely will these trainees actually be able to use this knowledge for their own programs (i'm open to counterexamples here). But the reward of going to work for say Google is so immense - both monetarily and in experience and connections, that students accept the cost of learning the wrong stuff for the future benefit, the job where you "actually start learning valuable knowledge"

The next level up in a career is looking for engineers with a proven history of being able to scale. Guess who that is: last year's hires, who got brought into an already working system.

This is the pattern we see in management of Fortune-500's and prestige banks. This has created a culture where you have to buy your way in the door with years of what a lot of people who actually do it, will admit is largely B.S. Again, this is not irrational though, because there is often a dramatic drop-off in quality of opportunities if you don't buy into the game.

So since everyone else is convinced (perhaps rightly so) that the real value doesn't come until you've been in an elite position, the effect is to erode all value in acquiring knowledge, skills and experience, in anything that doesn't neatly fit the filtering game.

[maverick_iceman]

>The stuff ends up being so esoteric, that rarely will these trainees actually be able to use this knowledge for their own programs (i'm open to counterexamples here).

In my own work I have to use the hyperloglog[1] algorithm pretty often. If I didn't know CS basics it would be a complete blackbox to me.

Or say someone is building a database and needs to make sure that chances of personally identifiable information getting released is minimized. He'll probably have to use some sort of differential privacy[2] algorithm. If someone has never encountered probability how would he even understand the requirements?

[1] https://en.wikipedia.org/wiki/HyperLogLog

[2] https://en.wikipedia.org/wiki/Differential_privacy#Formal_de...

[stillsut]

Thanks, interesting examples.

Could we drill into hyperloglog and how you use it? Is it a bioinformatics app? Otherwise why are you using this approximation technique on modern hardware?

[maverick_iceman]

Standard algorithms and data structures stuff are not that esoteric. I think companies are absolutely right in insisting that candidates know such basic topics like big O notation or various sorting algorithms. Nobody is asking them to go major in CS, it's perfectly possible to learn these things on one's own.

[stillsut]

As we know in math intense subjects, being familiar with != performs well on a contrived question, for example you'd probably fail your calculusII exam if you took it right now, at least I would. Even though you might have a better intuitive understanding of integration after having worked in the real world and forgotten the exact rules. Same for big-O factoids.