Comment by ascar | Hacker News Reader

Comment by ascar | Hacker News Reader

Parent Full thread

Your logic doesn't add up.

Sure, these tests are not an ideal proxy for job performance and indeed they shifted more and more from general intelligence assessment to algorithm coding assessment. Solving the algorithm is only part of the equation though. Writing readable code, explaining your thought process and decisions and communicating well with the interviewer are also vital points of the assessment. Is it a perfect or even good proxy for job performance? I dunno. Do we have a better one within a 40 minute interview? Not really. Past experience is unfortunately also a weak indicator, starting with the question if what's in the CV is an accurate reflection of the truth. There is also an interesting cultural aspect to lying on your CV.

Every approach comes with its own unique set of problems and as long as the position is sought after people will prepare and try to game the system.

The opinion of a recruiter on the matter isn't really the most relevant either. Especially Google should have quite some data on their various approaches and correlation to job performance and seems to use it. E.g. the trick questions were abandoned as they didn't show a correlation to job performance.

> these tests are not an ideal proxy for job performance > Past experience is unfortunately also a weak indicator

If past experience of doing the actual job is bad indicator and these test are not a good proxy, then according to your logic, experience is worthless and we should all optimize for these tests.

And this is exactly what is happening in the industry. Younger professionals spent hours memorizing Leetcode because they just graduate to pass the test. Older professionals having years of experience and family obligations don't.

> Writing readable code, explaining your thought process and decisions and communicating well with the interviewer are also vital points of the assessment.

Maybe to you, but I have seen many candidates get an automatic fail if they can't solve 2 questions in 50 minutes. The key metric being 2.

> Especially Google should have quite some data on their various approaches and correlation to job performance and seems to use it.

Source?

> Every approach comes with its own unique set of problems and as long as the position is sought after people will prepare and try to game the system.

Agree

The questions I see don’t use any tricky DS/algo that needs memorising. They are all solved by some loops and a simple data structure like an array or dictionary. If you can’t use those comfortably then you probably aren’t actually going to do well at the job. I don’t practice at all for interviews and have had no problem passing, I get an offer maybe 70% of the time.

If anything companies are desperate for experienced hires and give them really softball interviews. If you are a new grad you get grilled hard but if you are going for staff+ then you get asked some generic soft questions and a fizz buzz

(Have 15 years experience)

> The questions I see don’t use any tricky DS/algo that needs memorising.

That's kind of anecdotal. If it was that easy, everyone would pass. My experience has been entirely on the interviewer. The younger the interviewer, the more they have to prove by giving dynamic programming or direct acyclic graph questions. I am not denying your experience, but it just isn't universally true.

> If it was that easy, everyone would pass.

The implication is not that the questions are easy for the average person, but the questions are relatively easy to programmers talented in solving these kinds of questions (even if they don't actively grind problems on leetcode).

I personally think it's true that there might not be a high correlation between being good at solving leetcode style problems and being good at software development in general, and there's definitely a valid point to be made that some FAANG companies are essentially off feeding unconscionable ad money to a class of "smart" people who can leetcode without a bitter grind.

But when people complain about mandatory leetcode grinding, I just can't help but find it kind of ironic that they're simultaneously feeling bitter about not being able to get that $300k+ job, yet also feeling self-entitled because they have some so-called industry experience (which may or may not be relevant), and also proud of forgetting what they were taught in their CS curriculum... I suspect the Dunning–Kruger effect is strong in those cases.

At the end of the day, it's a job. If the employer wants you to grind LeetCode to get the job, then yes, you are going to have to do it. It still doesn't make it right though. I have no problem having to jump through hoops, but it does grind my gear when programmers use LeetCode as a badge of honour to show they are superior than others. Elitism within the LeetCode community is very strong.

> If past experience of doing the actual job is bad indicator and these test are not a good proxy, then according to your logic, experience is worthless and we should all optimize for these tests.

How do you assess and verify past experience of doing the actual job? That's the tough nut to crack and that's why they ask questions to assess your ability on the actual job. There are a lot of decent engineers out there with tons of experience and there is an equal number of below average engineers with tons of experience out there. How do you hire just the top 10%? As most work is done in teams how do you verify the candidate isn't selling his teammates accomplishments for his own while he contributed little of value?

> why they ask questions to assess your ability on the actual job

You already said LC isn't an ideal proxy to the job. So I am not sure what point you are making.

If I work at Stripe for 5 years as a software engineer, are you saying I am lying about my experience? How about Google? How about Coinbase? How did these people survive for so long?

It seems like you default to the candidate being liers. Maybe you are young, but it will happen to you one day, once you learn the full-stack and then realize no-one cares because you need to pass LC to even get a job. Something you don't need because software engineering encompasses much much more than some DSA algorithm.

Unless the previous employer is known to fire underperforming employees quickly, and willing to explicitly say so during reference checks, how does a prospective employer know whether a candidate's $n years in $company is a productive one or not?

So you haven't seen underperforming people survive at large famous companies for multiple years? Are you sure you have that much experience in $bigco? How old are you really, since you say the GP is "young"? I wouldn't be so patronizing to others if I were you.

> I wouldn't be so patronizing to others if I were you.

Perhaps you are seeing it through the wrong lens. I am simply stating that coding is much more just DSA.

> How old are you really

If you don't know how rude you are, then I don't think we should continue this conversation. It seems like you do not know how to respond to someone you disagree with using manners. You are entitled to your own opinion as I am. Being confrontational and asking someone's age outright doesn't make your case and tells me you lack basic social skills.

Well obviously it's not working well, I've meet better engineers in the defense and automotive industries than I have at google -- which to be honest is better for everyone when you think about it

Reply on news.ycombinator.com