Algorithms In Interviews SUCK | Prime Reacts

Recorded live on twitch, GET IN

MY MAIN YT CHANNEL: Has well edited engineering videos

Discord

Hey I am sponsored by Turso, an edge database. I think they are pretty neet. Give them a try for free and if you want you can get a decent amount off (the free tier is the best (better than planetscale or any other))

[JensRoland]

Prime: “Algo interviews show that you’re wise and skilled enough to practice algos for weeks to show that you can learn something hard and won’t fight the system”

Also Prime: “If Google wants to offer L69 but I have to spend two hours on a take-home test, TAKE A HIKE GOOGLE, I got four kids to name and a beautiful wife to defrag!”

And there you see the problem with algo interviews: the devs who can’t be bothered to waste weeks doing leetcode aren’t always stupid or unwise contrarians; sometimes they’re busy experienced people who value their time and would rather spend those leetcode weeks on real-world valuable stuff like learning Go or burning React books.

[HeaDStrongTerran]

In my experience it just depends on the interviewer. Maybe that is why people fear them. I had interviews where I was asked to solve rather involved problems and the interviewer was responding to questions, was giving hints to me missing things and leading me along. So I could really demonstrate my skills. And then I had interviews where the interviewer asks me some trivial maybe bit gotcha question on some code snippets and then gives me a complete silent treatment. Some of them I did not immediately catch and see the problem - zero feedback. Some I explained in detail the problem and the solution only to be met with - so you think this code is good or bad? I completed the interview and immediately sent them an email that I am withdrawing my application. Remember when you interview with a company it is a two way street. You are there also to evaluate them and if you have the luxury you can dismiss weak candidate employers.

[RicardoSilvaTripcall]

Most of the people just grind those subjects 3 months prior to the interview, today it's really clear that tech interviews are a different subject that you have to spend sometime preparing to, and like most of the test that you have to prepare to take, you will forget almost everything less than a month after, dunno to me, tech interviews has gone awry, it's a complex take nowadays ...

[etagh]

You are saying that you don't have time for take home, but how is that different to algorithms, where you need to invest also significant time at home to prepare for algorithms?

[yaksher]

I feel like the problem with leaning too much on algorithms in interviews is that it just becomes a "did you put in the practice time to memorize enough algorithms patterns in typical interview questions" which doesn't really measure any particular ability to program. The problem is the same as the problem with every other heuristic: the moment you let people know what it is, they can game it. It's part of the reason colleges use such complex selection criteria and why YouTube obfuscates its recommendation algorithm: the moment you tell people what it is, it falls apart, because it's a heuristic and people can always game it.

Algorithms interview type question are a useful proxy for knowledge _if they're not the standard thing everyone does, _ at which point their value becomes greatly diminished and they're no longer testing much except whether you practiced for them.

[maciejcisowski]

My experience: it doesn't matter if you're asking directly about alogs that much. It's important how you approach the answer part. If you just wait until the interviewee gets it right or dies from stress, you're doing it wrong. If you're more interested in having them show you their thought process and fluidity in working with a given programming language, that's great. Let them think out loud, see how they react to hints and if they can gauge the efficacy of their own solution. Let them course-correct if they get it wrong - you'll get to see how they debug and how they react to failure. Don't just throw linked lists at them, they won't ever use them.

[-Jason-L]

I have worked as a dev at many companies over 30 years. Including a large silicon valley one. I have never once had to deal with DSA in an interview.

[kennythegamer]

Knowing algorithms isn't a secret handshake if the company never responds to your application.

[IgorGuerrero]

I love how Prime thinks he doesn't gate keep but he's right now doing that.

There's a lot of pressure interviewing, because your livelihood depends on it, also you just don't know how introverted somebody can be. I think is funny we see these extroverted figures in social media thinking they're normal, but no, a programmer that can do Youtube is not the "normal", obviously somebody who streams every day in front of thousands of people don't get any pressure when whiteboarding... You don't say...

[etagh]

Also my problem with algorithms challenges is that they are often very time constrained which adds additional pressure / stress which you will never experience on a job.

[mosesnandi]

Alright just to throw this out there...this guy has a full 9-hour course on Algorithms!

[ErazerPT]

Kind of on the fence here... Crux of the point being, whatever "deep ish" we're talking about, either a) someone WAY BETTER than me has already solved it and/or b) it's WAY ABOVE my skill set. So, if stuff is at "that level" I'm just the wrong candidate for that job no matter what...
Practical example base on an internship i did a few years back, i can't write any paper in Arxiv regarding ML or come up with the corresponding models, i can't write Python/Tensorflow/Keras from scratch and my high school math surely isn't up to writing gradient descent. But none of that stands in the way of being capable of bringing it all together and pushing out a prototype that met the client requirements.
Just because you can't make an internal combustion engine doesn't mean you can't drive a car, and just because you can doesn't make you a good driver. Horses for courses...

[zoeherriot]

As someone with autism... yes, it's extreme pressure. I get by, I can do an interview. But it's harder than it needs to be.

[xDELTAGIx]

The problem with these interviews is the interviewers. I did my algo for Amazon in Rust and he couldn't read the code so I was failed lol, but since then I started my own project and im at 200k lines of rust so suck on that Amazon

[asagiai]

By schema builder do you mean like;
some sort of json encode/decode because I think that is the easiest way to create object?
or do you mean something like a function that takes an object and a new property, and you need to insert that new property to the object returns the new object with that property?

I think asking someone to make some sort of schema builder is a better way to test someone's programming knowledge or experience. But technically this is not an algorithmic interview.
This is more like knowing and implementing DS

So my take is
1.) I think there is a disconnect between what you are thinking and the blogger. The blogger is more like don't let someone knowledge or lack of knowledge about algorithms becomes the deciding factor to hire someone (example if someone don't know how to implement Dijkstra's algorithm). Instead, it should be something they learn on their free time or something to improve one's self.
2.) A) Is it okay to ask for algorithmic style interview. I would say Yes. But limit it to what will actually be used.
B.) A person's knowledge about DS and its inner working are more important in most cases.
C.) Other technical questions. (example asking someone if they can implement a recursive function based on a changing condition)

PS. I'm always thinking about it, but I just confirmed today that quicksort is in-place sorting algo.

[autogenes]

Can somebody explain what's wrong with copilot's quicksort? Is it the fact that it kept pushing onto new arrays instead of swapping values?

[theohallenius]

Just had one, terrible. You can't expect someone to invent a way to clear (zero out) every element in an array memory in ideally near constant time on a shitty virtual doodleboard in 30 minutes. Maybe I'm dumb but this sort of stuff is a huge waste of time during an interview and not very useful to test overall abilities. Even if you have a clever idea of solving something that hasn't been solved for decades it it's difficult to illustrate let alone implement.

[rocknowradio]

Algorithms interview are like assessing in chess how good is my Alekhin defense (which I never played, never will) when I'm playing Sicilian and French. This is not development, is math.

[boot-strapper]

I've tried to spend time leet coding. But I hate doing it, and im not good at it. But I've been successfully employed in fortune 100s for 8 years. So from my personal experience I just dont think its important.

[tHebUm]

20:25 My current employer does take-homes. Then two interviews: one with the hiring manager (~one hour) and one technical (~hour and a half). Much preferred it over my Amazon interview process which was an online timed test followed by 6.5 hours straight of interviewing.