Hacker News Reader

Comment by cmrdporcupine | Hacker News Reader

Having a CS degree gets you through a Google-style interview -- where they will hammer you for hours on your ability to recall the skills you needed to pass your algorithms classes in university, and grade you almost solely on that.

And then you will start working there and almost never use those skills again, because you'll be using a vast library of datastructures and algorithms in the language of your choice, and the skill you will need the most is the ability to analyze why you might need one vs the other.

Unless you're on the team which is writing said libraries, which is rare.

View on HN

Implementation is only a portion of what's important in an interview. The thing you will not get anywhere without is a knowledge of what algorithms exist and when to use them.

You will not get through a Google style interview knowing what algorithms exist. You will be expected to implement them. On a whiteboard. Almost perfectly. Without access to a computer.

"Only a portion". And almost-perfectly is a big deal. You're not expected to write compiler-ready code, just to show that you understand what's happening under the hood. Which meshes up with practical expectations -- if you can't understand what potential bugs/space/time complexity you're introducing when leveraging a library, then you're a fundamentally worse coder than someone who can.

Except this is in fact bullshit since if you're ever working on some speed/memory critical code, that's the time to bother filling your brain with the different options available and picking the appropriate one. Carrying around arbitrary knowledge to show off is pointless.

Most code most programmers write won't ever run at sufficient scale or be sufficiently critical that CPU or memory bounds will matter, and using e.g. one sorting algorithm vs another will be largely irrelevant. It's recognising when you need to bother optimising that is the sign of a competent developer.

And asking someone to write actual (not pseudo) code on paper/whiteboards - as I've seen in interviews - is like assessing someone's driving ability by asking them to mime driving a car.

I agree.

But it's worth pointing out that a Google style interview also includes some portion which is "system design" oriented which would touch more on that kind of thing.

But they're more interested in avoiding false positives, than false negatives. So the process is designed to weed out most people.

Reply on news.ycombinator.com