Hacker News Reader

There's possibly some worldview mismatch on your comment, because leetcode isn't "basic scales" programming for any specialty.

It's much closer to expect your world famous violin player to sing on the stage than it is to expect them to play scales on the violin.

But then, maybe not. Some people call all kinds of random exercises by "leetcode". The author may be one of those, I don't know if anybody can know this.

Basic data structures and algorithms knowledge is a base skill that will translate to better success in any software job. Yes you may not be coding graphs in your day job (although you might - I have personally done it before), but someone who can grasp stuff like BFS vs DFS and Dijkstra's algorithm will absolutely be better at the job than someone who comes in and says "idk who cares, I'll just use a library".

BFS? Really, there are libraries for that and I would hire a kid who timidly used one as they could not remember how to implement it over someone who confidently introduced and off by one bug because they thought they knew how to implement it without references.

BFS/DFS is astonishingly simple. It’s like 10 lines of code. And it’s almost always done (in practice) with a custom type where you work your way through custom fields pointing to the next vertex, so it’s not an algorithm / structure that lends itself well to standard libraries. It’s not like a self-balancing tree (which is actually a bug-prone, large implementation that you should just let the standard lib handle).

If you can’t trust the people sitting next to you implement a simple graph traversal without bugs, then your organization has a professional standards problem.

> BFS vs DFS

Yet, on the real world the answer is always one of "either" or "neither".

Anyway, most leetcode aren't of the type of "a widely applicable algorithm solves this". People only know the algorithms that apply to them because they studied those problems specifically.

And notice that one of the algorithms on your post is named after a researcher from the 80s. That's some 40 years after computer science became a thing. That's how long it took the entire humanity for learning it without anybody specifically telling it to them.

Absolutely not lmao. Many problems are more efficiently solved with bfs vs dfs or vice versa. If the problem requires you to reach a leaf dfs is usually better

What? Neither of them are efficient in any practical sense.

Once I started interview practice on leetcode I began to see algorithms & DS in everyday coding tasks

There's an enormous gap between knowing when to use those tools and creating a solution with those tools when you already have the job, vs banging out a perfect solution in 20-40 minutes with someone looking over your shoulder, who is also deciding whether you deserve a paycheck based on those 20 minutes.

Not to mention that the interview process in a lot of fields without any objective hiring criteria consists of the hiring manager being friends with your uncle or your grandparents going to the same country club as the CEO. I'll take what we have, thanks.

Author's point is not that it's too simple but that it is not relevant. It's not similar to asking a mathematician to do first year math problems, it's asking them to, dunno, remember obscure regulations on accounting.

You got it backwards. It's like asking an accountant to do basic math. I don't trust you to handle my money if you can't even do that.

This tweet is about "grinding", eg the time-consuming prep work before interviews. I think.

I'm a prof. I grind when I prepare a lecture. It's part of the job.

Grind? Really? I work in higher ed and I teach. Prepping even a new course is not grinding. Teaching a course where you are professor of record is certainly not grinding unless you dont know the topic at all. If that is the case though you should not be lecturing for this class.

The only legit excuse for grinding here is if you did not do any prep.

Doing your job even with a 4/4 load is not grinding.

Research proposals and research projects might require grinding. Prepping lectures no way.

I hope neither I nor anyone I care about ever ends up in any class you teach…

Unnecessary attack. You disagree on what "grind" refers to and that's fine, but your comment was personal and mean.

really? how about if I wrote this? not personal or mean, yes?

“Teaching a course where you are professor of record is certainly not grinding unless you dont know the topic at all.”

the absolute best professors on this planet are the grinders… if you are not, you are just coasting on something you think you know but you really don’t (probably re-using powerpoint presentations you created in 2012…)

Clearly I touched a nerve. Perhaps you view grinding as "doing your job". I happen to have a different definition.

Originating a class can be a slog and originating two to lecture next term would be a grind. If you are not professor of record and thus likely did not originate the class then your assignments and reading, hell the entire course is likely layed out for you. Teaching a section of a course you originated or that someone else originated is likely 80% of the teaching load. Familiarizing yourself with this material so you can effectively teach it is doing your job, not grinding.

I'm not sure why you feel the need to suggest I am a crappy teacher or that I lecture from notes more than a decade old. Knowing how accreditation works, how courses are originated, and how sections of thoses courses are taught at r1 institutions, I can only conclude you are ignorant as to how these things actually work.

As a real world example, there are many great online courses from places like MIT and Stanford. Hell you can sometimes even get these online lectures for free, the same material covered by those institutions as part of their degree programs.

How much grinding does it take for a lecturer to to proctor such a class, particularly given that they also likely get FAs to do labs and grading?

If you attended or work at a college where professors grind out new classes every term then that institution is going to have an accreditation and/or tenure issue as "special topics" dont make a degree program.

For example, from the perspective of an algorithms junkie, there's literally 0 way an algorithms professor with algorithms research published would forget any of the basic algorithms they teach in an introduction to algorithms class.

Now what may happen is forgetting what it's like to be a naive, 18 year old freshman in college, and not somebody embroiled in academia for the past few decades.

But the audacity to suggest that it's even remotely within the norm for professors to not know their own material... wow.

As I said prep is grinding.

All other serious endeavors require endless “grinding”. Imagine a hockey player being annoyed about being asked to tryout. Imagine whining about having to practice their skating or shooting.

Compared to many other things, software engineering requires astonishingly low prep. Your average high school baseball player has spent more time in the batting cage to earn a varsity letter than the median SF SWE has spent on leetcode

Batting in the batting cage is much closer to skills required to perform as a batter than Leetcode problems are to perform at most jobs as a software engineer.

Is there some overlap of Leetcode problems and real world problem solving? Yeah, a little but most of my time is not solving yet another algorithmic problem but thinking about the problem from a business/product perspective, decomposing it, uncovering and understanding requirements, designing code that will be easily read and modifiable by people in the future, etc. 95+% of my time on the job is not trying to figure out what's the best way to find the number of inversions of a red-black tree.

There are jobs that hard and deep knowledge of algorithms is important, for those please go ahead and keep testing people on it. The vast majority of the jobs requiring people to jump through Leetcode hoops is not that, and a lot of those interviews are not testing for what is actually needed to perform the job well.

> If you think something you have to do in an interview is “humiliating” when all they’re asking you to do is demonstrate some skill on a whiteboard you have some self-reflection to do.

I don't know about the author, but some places are asking for a coding exam that takes hours or days. The process of application itself can be a long and complex ordeal where they suddenly stop responding, only to find out weeks later only after you chase them that they already had somebody for the part but wasted your time to make it seem like they went through a proper process. Then if you get the job, you move to the area and sign up to a 6-12 month rent agreement when the probation period is 3 months or less where they can dismiss for any reason.

I think more can be done to even up the relationship. The more the employer demands as part of their process for hiring, the more they should provide the person going through these hoops. The risk is really quite unbalanced.

I know excellent engineers essentially stuck where they work because they cannot afford to miss the mortgage payments. Yes they can get interviews all day long, but the risk of jumping to a different ship is immense.

I like leetcode, and I think the practice is defensible. To be fair, I've made a lot of mental loops to get there.

However, I'm also sympathetic to the "leetcode is useless because you don't do it in your day to day" camp.

I used to lean more towards my mental loop side. But nowadays there has been a big question due to a recent job interview I did.

I recently applied to become a data analyst at a F500 company. It seemed fun and the pay was marginally better than what my software engineering career has been (as I'm not from the US).

I remember that when I applied, I didn't even know what a data analyst was. However, they gave me a technical assessment and I aced it.

Zero prep

Completely aced it

They told me that their work sample round was indicative of the job, so if I could do that, then according to them: I could do the job. It took me 4 hours in total.

Of course, there are tech companies that do this. However, I was a bit stunned that this was so easy. To be fair, getting in wasn't easy as I applied to some data analyst positions for fun to other companies and they all mentioned "you have 0 years of experience in this role". So I was lucky in that sense, and I was also lucky that they were willing to look at the person as opposed to treating me as a number.

There have been 2 tech interviews so far that went the same way. With that I mean: 0 prep, just using your skills and you got in based on those skills.

Look, I like leetcode, but after my recent experience I'm just not sure if it's the best way to assess talent. One could make the argument that it's an easier role? But I don't think it is, I find CRUD relatively easy too which has been most of my job. Though, there was this one week where my leetcode skills really helped as the company had a genuine novel leetcode hard style type of question.

Did you get the Data Analyst job ?

Agree with everything but I don't know think James Maynard was actually being interviewed during that video.

True although I get the feeling from James Maynard that he’s the sort of person you would almost have to physically restrain if he was in a room with a whiteboard in it and you didn’t want him to solve some problem on it.

Demonstrating core competency is one thing, leetcode is typically not that though.

I completely disagree with you seanhunter ( and everyone else mostly) so please dont take offence I could be wrong but the following is my opinion.

I am thankful most companies are gatekept by people who think like you as I feel it gives me and my company an advantage. In the UK over the course of nearly 20 years I was asked by random interviews to do "fizzbuzz" which I always refused to do out of principle.

The problem being anyone who doesn't know how to code could just memorize it.

My problem was it wasn't a "starter" question it was the equivalent of asking me if I wear shoes on a daily basis and if so, to demonstrate it in an extremely concise way.

I felt it was offensive regarding my ability, indicative of narrow minded testing protocols for candidates, and a lack of respect for why they had chosen to interview me as it meant they probably were unaware of my portfolio, which many companies were.

I am not saying this to convince anyone on HN they are a bad idea as most of them rely on codemonkey exercises like these to prove their worth straight out of college, I am just saying I think they are idiotic.

Many people fail fizzbuzz. Most interview questions at big tech are not much harder than fizzbuzz. You are not qualified to be a software engineer if you can't write fizzbuzz after seeing it for the first time in your life.

"Many people" is an understatement. In my experience 70%+ of applicants fail basic fizzbuzz or reverse a string level questions. If you think these questions are beneath you, good, you've just made my job easier as an interviewer.

No, you’ve just optimized for a lazy metric. You’re filtering for people who can handle trivia under pressure, not necessarily good engineers.

If you are applying for a job in the front office of an NBA franchise, and I ask you "which team did Michael Jordan play for?", answering with "this is a useless trivia question, it is beneath me, I can easily google it, how does it matter?" will mean you are not going any further in the interview. You can then complain about it online all you want, but that's how it is.

Similarly, if I ask you "write a program in your language of choice that prints all even numbers from 1 to 100", you can either cry about it or take the 3 minutes and actually do it. If you can't, it's a pretty clear signal that you don't actually write code in your day job and aren't suited for this one.

It's not trivia, it's basic problem solving skills of the trade.

Why would someone be under pressure when answering "basic fizzbuzz or reverse a string level questions"? Personally I like easy and quick warmup questions, they make me more calm

This was one of the most eye-opening truths of the industry revealed to me when I started interviewing candidates. There are so many software developers, even with long histories of well-known companies on their resume, who just cannot write code. Even Fizzbuzz. Even "write a for loop with an if statement" type problems. Under pressure or under no pressure. On a whiteboard or on a computer. This has been shockingly true about a pretty consistent percentage of candidates for every company I've worked with as an interviewer for 25 years.

This is why we have leetcode and Fizzbuzz. There are simply too many phonies applying and counting on their charm and charisma to get them the job, which sometimes works when you don't make candidates actually demonstrate their skills.

FizzBuzz trips people up for all kinds of reasons: fear of ‘gotcha’ questions, nerves, overthinking, or even feeling insulted by how basic it is. Sure, some fail because they can’t program—but passing doesn’t mean much either. It just means they’re good at FizzBuzz, not necessarily at solving real problems.

Sure, some qualified people fail simple problems. More likely the applicant failed because they didn't pay attention in school and made a career of copy-pasting stuff from stackoverflow(or AI these days).

I’m not taking offense but don’t think for a moment that you know me. I don’t “gatekeep” and don’t in fact ask leetcode questions when I interview. But I have been a software engineer for 20+ years in every possible role from a basic IC up to CTO, been part of small teams, run huge teams been a cog in a massive machine to being “the guy who has to do everything” and I am perfectly happy to do leetcode in an interview and wouldn’t consider it “humiliating” and what I’m saying here is if you think it’s offensive you should adjust your perspective. I personally don’t think it’s great signal in an interview for some of the reasons you say which is why I don’t ask these questions myself.

I don't think I said I know you. Cheers for your reply , maybe my perspective is wrong.

If I were given fizzbuzz as an interview question, I would say something like: "Sure, I'll do it. But I've got 40 years of experience, and I expect the rest of the interview questions to be on a bit different level."