Hacker News Reader

Comment by atleastoptimal | Hacker News Reader

I am probably in the minority but I actually like Leetcode style interviews and to a lesser extent system design interviews. I think they stick to this style, even if it isn't ideal, because

1. Easier to judge general capacity to break down problems/think on a smaller problem that doesn't require as much prior knowledge

2. Easier to distribute the range of people who can conduct interviews, rather than needing a specifically competent senior engineer whose time is more valuable

3. Consistent style and non-leading interviewers make it easier to compare interviewees on a case-by-case basis rather than there being a risk that certain interviewers style lends candidates performing better/worse

As bad as it is, one good interview round and you're set for a very high paying job for the next 5+ years at a company with great benefits. Even with the shittiest interviews I think it's a good trade off.

View on HN

leetcode does not test for general capacity or whether you are able to build anything

it just tests how fast you can reproduce an optimal solution to some problem that for some non sensical reason became part of a body of coding challenges

you are wrong in thinking that leetcode is a fair interview style. it gives people who happen to do well in the leetcode style an unfair advantage over those who do not or have no time to repetitively memorize the problems and solutions.

When I mean fair I mean less biased based on prior domain-specific knowlede. Of course people who are better at any X-style interview will be advantaged over people less good at X-style interviews. However I feel that if it were really useless then it wouldn't be the status quo for so long for so many billion-dollar companies who easily have enough money to put more finesse into hiring if they wanted.

General capacity over whether you are able to build anything is harder to intuit in interviews, but resume, past experience, systems design style interviews, github/portfolio give a good estimation.

it is the status quo because it allows companies to discard the value of experience because leetcode style interviews do not take that into account. once you are hired they can profit from your experience without really paying for it. it also lets them avoid potential discrimination issues. always ask yourself what the real reasons are behind the “status quo”.

I don’t hire, but if I was, then for finance I’d hire a guy with finance and stock exchange automation bias, who knows e.g. common pitfalls in working with exchanges. So they could avoid losing money or creating an accounting mess.

For business I’d look for a guy who dealt with bigcorp bullshit protocols and e.g. knows up front that “there’s API” may mean anything from exactly that to “we have a half-working semi-email system with rate-limited polling and no idempotency”. So they could make correct predictions and decisions.

For both I’d like to see 5-10kloc past or example project, or at least few pages of code that they assume is both practical and good enough, to get the sense of their development (not CS) skills and habits.

I have no idea what for I’d hire if they excel at leetcode. That position must probably relate to some rocket science in a low-power low-tech environment where external dependencies are forbidden by a sort of a military contract. Because irl when you need btree/redblack/matrix/etc, you need postgresql or import <package_name>. Not to write it from scratch because you smart.

if it were really useless then it wouldn't be the status quo for so long for so many billion-dollar companies who easily have enough money

Their money can what others can not. But that doesn’t mean that all others aren’t just practicing cargo cultism.

Are you going to die practicing? Is this going to make you a worse developer?

There is like literally 3-5 patterns that you can compose together to solve most of the problems.

To clarify: I agree that most of interview processes are shit. I disagree with the tool class like leetcode being called shit.

> Are you going to die practicing?

Probably not, but it's bullshit in the sense that solving leetcode problems on the fly is not typical day-to-day programming work in ~99.9% of jobs. And even when creating a complex algorithm is a part of a job, it's usually for a big project that's the core of some business problem that you will have a considerably longer amount of time to research, think about, test different solutions, and implement than the space of video coding interview.

Would you hire a divorce lawyer by testing them about the intricacies of Constitutional law? Would you hire a heart surgeon by quizzing them on the complications of laser eye surgery? Would you hire a civil engineer to build bridges that support trucks and trains by asking them to solve some underwater pressure equation involving submarines?

> Is this going to make you a worse developer?

In a certain sense, yes, because there's an opportunity cost to every human activity. If you spend 500 solid hours of intense effort grinding out leetcode solutions, that's 500 hours you don't get to put into practical stuff you might use on a day to day basis: learning about servers and web services, deployment, different software architectures, databases, design patterns, security, accessibility, etc.

> There is like literally 3-5 patterns that you can compose together to solve most of the problems.

This is somewhat misleading to me.

You can learn some of the common elements of solutions like trying to break up a problem into smaller parts, using a hash table, sorting an array, recursion, or other common elements to leetcode style solutions. But solving a type of problem you haven't seen before in the span of a video interview is pretty much a crapshoot, particularly when you have somebody just staring at you and typical problem-solving approaches you would take like doing Internet research or trying to diagram a solution out on paper are probably out of the window.

Leet coding is testing two things:

- Intelligence. It's basicall a form of IQ test.

- Willingness to prepare for an interview (by grinding out the leet code in your free time). This is a good proxy for general work ethic, and also for much hard you actually want the job.

it does not test intelligence, just whether you memorized a solution to a problem

hiring someone who is really desperate for a job is a poor strategy for those hiring unless the goal is simply to get cheap labour.

Without the pressure/context of a job interview, I find Leetcode to be immensely fun :)

However, within the context of a job interview I do unfortunately find it to be a little insulting. I think, like other commenters have mentioned, these Leetcode style questions are only to see "will you tapdance" - which is totally fine.

You're basically halfway there, perspective wise. Just leave your attitude at the door and do the interview.

For better or worse there really isn't a lot a company can do to fairly evaluate if someone can code other than actually do a little coding. And any coding problem that can be reduced to a 30-45 minute block is going to be somewhat arbitrary and leetcode-ish.

Good luck on your search!

Reply on news.ycombinator.com