If I am interviewing someone and if they ask me if they can google X, if it is easily googlable, I would just tell them what it is. Or let them google it. But I do not expect them to spend minutes googling it.
It is really unlikely, given a leetcode hard level problem to solve within 25-35 minutes and you can google your way out of it unless you are googling for the solution.
Isn't the entire point of memoizing this to demonstrate you are able to study/grind rather than prove intelligence? Different companies/managers have give me different reasons for the usage of this style of interview but it was usually to 'assess ability to get through rough stuff' and 'assess cognitive abilities'.
Idk anymore either and I refuse to conduct these interviews.
I generally prefer somewhat realistic problems that are multi-disciplenary but have ties to real world applications related to the role.
Leetcode style interviews probably serve two functions:
1) A way to suppress wages and job mobility for SWE. Who wants to switch jobs when it means studying for a month or two? Also, if you get unlucky and some try hard drops an atomic LC hard bomb on you now you have an entire company you can no longer apply to for a year.
2) A way to mask bias in the process while claiming that it’s a fair process because everyone has a clear/similar objective.
Meet someone who went to your Alma mater? Same gender? Same race? Give them the same question as everyone else, but hint them through it, ignore some syntax errors, and give them a strong hire for “communication” when they didn’t even implement the optimal approach…
Or is it someone you don’t like for X reason? Drop a leetcode hard on them and send them packing and just remain silent the entire interview.
To the company this is acceptable noise, but to the individual, this is costing us 100s of thousands of dollars, because there’s only a handful of companies that pay well and they all have the same interview process. Failing 3 interviews probably means you’re now out $200-300k of additional compensation from the top paying companies.
I’ve interviewed for and at FAANGs. I can’t believe the low bar of people that we’ve hired, while simultaneously seeing insane ridiculous quad tree/number theory type questions that have caused other great engineers to miss out on good opportunities.
Someone will reply to me “if you know how to problem solve you will always pass.” Ok, come interview with me and I will ask you verbatim one of those quad tree/number theory/inclusion exclusion principle questions and I’d love to see you squirm, meanwhile another candidate is asked a basic hash map question.
I'm sure anyone determined to do so can act unfairly regardless of what process is in place, but the fact that there is a standardized test in my mind does the opposite and makes the process much fairer. Assuming a fair-minded interviewer, the process gives a chance to a candidate whose resume may have less vaunted names on it to demonstrate their skill. I'm quite sure that I'd never have had some of the opportunities I have, not having a CS degree, if it weren't for whiteboarding interviews. I can't imagine any possible process that would thwart interviewers intentionally subverting it to hire their friends.
So are we just going with a base assumption that interviewers can NEVER be trusted with anti-bias training and learning how evaluate people fairly? The examples mentioned in this comment section are all blatantly intentional biases that people are choosing to use. The amazing part is that all the “standard test eliminates bias” people seem to the most ignorant to where bias helps THEM. Forcing people to study for two months is blatantly discriminatory to age and family status, at a systemic level. While “this white straight guy might explicitly choose to give the other white straight guy an easy question,” is very subjective and intentional on the individual level. Like, employees can always choose to do bad things, in any situation. That’s why we have at-will employment…
Is the culture just so broken at these companies that it’s hopeless to expect people NOT to blatantly exploit the system for their friends? Why don’t people get fired for doing that?
> Forcing people to study for two months is blatantly discriminatory to age and family status, at a systemic level.
How is this any less discriminatory than any other assessment based interview where you need to prepare? Non-assessment based interviews end up being vibes based which is much more discriminatory.
You’ve set up a false dichotomy here; for any given position, there is typically a way to conduct an assessment-based interview that doesn’t require too much preparation for qualified candidates.
For example, at my current job, we hire web developers with Rails experience. Our technical interview process consists of either a pair programming session or an async/take-home task (candidate’s choice) which requires the candidate to implement a small feature in a Rails codebase. We do have some room to improve on objective evaluation of the candidate’s performance, but there is a test suite and a rubric which we use to evaluate their work. None of this should require that the candidate study, unless they’re coming in to the interview without Rails experience.
This may work out great if you happen to have worked on Rails for your last job. However, I doubt that everyone you interview is actually that familiar with Rails but rather is pursuing any sort of opportunity that they can get. In that case, they would actually more time to brush up multiple different tech stacks than simply on algorithms.
In that case, the interview question is working as intended. For most of our roles, we want several years of Rails experience, and we are clear about that fact in the job listing. If someone applies without Rails experience, they either didn't read the job listing, or are desperate to find any job. While I empathize with folks in the latter situation, our positions really do require the experience, and the job market isn't so bad right now that a smart candidate should be going a long time without finding something.
If you happen to have Rails experience but it was several jobs ago, the task we give you is basic enough that you should be able to Google what you need during the task to refresh your memory fairly quickly. In fact, I did this when I applied, having not worked with Rails in a number of years.
Edit: My main point is, even if you technically do have to "study" (really, just Google a few basic Rails concepts) if you're rusty, everything you do is preparing you for the actual job. Studying how to implement a hashmap, or computing the longest palindrome from a string of characters, or whatever other harebrained problem FAANG etc want to ask, is 99% of the time not really helping you prepare for those jobs.
No, it doesn’t. I was just giving an example from personal experience.
For any given job, I believe there’s a way to design an assessment that allows the candidate to demonstrate skills directly relevant to the job, even if the job doesn’t have requirements that are as specific as “Rails developer with a few years of experience.” I have yet to see a single job listing in several decades in the industry where I think to myself “ah yes, the only way to find qualified candidates for the job would be to ask them to implement a quadtree.”
I've done a lot of whiteboarding interviews in my life and it's never been "implement a quadtree." It's always, at its core, a small coding probably that could plausibly be part of a real-world problem, and therefore it is directly relevant to the job.
I've also never been asked a question like this, but it's because I specifically avoid the companies that are known to ask them. Sounds like you have as well (knowingly or not). But the OP is about LeetCode-style interviews, so that's what I was discussing.
Are you claiming that no company asks irrelevant leetcode-style questions as part of its interview process? If so, that’ll require some more evidence than “well, I haven’t been asked any such questions.”
I'm rejecting your characterization of "leetcode-style" questions as 1) irrelevant 2) primarily about simply implementing some obscure data structure (as opposed to correctly using data structures).
I'm sorry if I wasn't clear; I didn't mean to say that all LC questions are irrelevant. I was just asking if you're claiming that no company uses LC questions which are irrelevant. Because if you are, that seems like a pretty difficult claim to prove given all the stories out there on every company review/discussion site like Blind or Glassdoor.
And my "implement a quadtree" example was hyperbole; again I'm sorry if you didn't recognize it as such. I do suspect we have wildly different definitons of what questions qualify as "relevant to the position," though, so I'm not sure how far we're going to get in this discussion.
Because the interviewers themselves typically have wide latitude to select the actual question, no, I don't think such a thing can be claimed. But I'd say the majority of questions asked are reasonable ones that can plausibly be solved in the timeframe and the majority of interviewers I've observed from that side of the table have not expected perfection or wanted their candidates to fail.
I don't think I (nor most others commenting here) were claiming that the questions couldn't be solved in the timeframe, that interviewers expected perfection, or that they wanted candidates to fail. So I'm not really sure why you brought that up.
>any other assessment based interview where you need to prepare?
because most other assessment based interviews are based on what you do on the job, likely stuff you've done at previous jobs as well. Less prep time when you spent thousands of hours already as a career.
>are we just going with a base assumption that interviewers can NEVER be trusted with anti-bias training and learning how evaluate people fairly?
Yes. Because interviewing is
1. hard, but no company has proper full time proctors. So "expert interviewers" are a rarity
2. not standardized in the slightest. So your performance varies entirely by the interviewer, their style, and their mood that day.
3. some weird blind test where you hope you studied the right topics. You're not getting the best out of a candidate, you're basically hoping they read your mind and know exactly what you want them to say.
>Is the culture just so broken at these companies that it’s hopeless to expect people NOT to blatantly exploit the system for their friends? Why don’t people get fired for doing that?
Sure. but that's a universal problem. "It's who you know, not what you know" is advice that has spanned centuries. I can't even blame the modern tech industry for that one.
That + the above issues with interviewing mean you're always going to go with a referral over a random applicant.
> 2. not standardized in the slightest. So your performance varies entirely by the interviewer, their style, and their mood that day.
That isn't always true. When I interview HR gives me the exact questions I'm allowed to ask. These are vetted both to prevent me from asking something illegal, and also by research to get the type of things useful for interviews. Sometimes it is annoying - you can easially finish the interview with a great score but I have no clue if you can write code or not. However we are carefully trained on how to ask the questions and how to grade them.
That makes sense for soft questions. But I doubt it's HR devevloping a dynamic programming problem and writing a rubric for how good a score you can give based on a response.
I was mostly referring to technical tests, but I understand there are definitely some set of questions you need to ask no matter what. I don't really knock recruiters too much for repeating the usual "are you authorized to work in the US" kinda stuff even though it is the first question on their job application.
1. I’ve worked at multiple companies that absolutely have expert interviewers who design the interview questions and then teach mid-level engineers how to proctor them correctly. It’s like one 30 minute meeting, it’s not that big a deal.
2. All tech companies I’ve worked at since around 2015 have completely standardized interview questions, sometimes also hosted in a GitHub repo where any employee is free to comment or even open a PR to request a change. Every candidate gets the same questions. This thing where “faang” interviewers just pull a random LC question out of their ass is complete insanity to me. An organized set of questions takes a senior engineer like a week to organize and commit to… And if your questions can be memorized and recited by rote memory and the candidate can do well on it without the proctor knowing, then your questions are BAD or your proctor is a moron.
3. I’ve never worked anywhere where I would describe the interview like this. Even the startup where I was the first engineer and there was no formal “test”, it was an hour long chat with the technical cofounder where he grilled me about coding skills and past experience/accomplishments. I won’t take a job if the interview isn’t asking me things that focus on my existing experience and skills, it’s a red flag about the company culture in general.
As for your “universal problem,” I disagree with fatalistic takes where you just throw your hands up and say “whatever shall we do” all the while YOU are the one benefitting from the system that cannot be changed. This is how simple-minded people think about the world.
1. I've 100% met with interviewers that felt like they were googling interview questions on the fly. YMMV which was the point of the "no standardization" argument. I've had interviews with no whiteboarding, some with whiteboarding, and some that simply felt like we both wasted our time.
2&3. I wish I could relate. I work in games, fwiw, so maybe it's simply more wild west than the other domains in such regardss. It's not even that I can't answer some of the questions, but if I'm asked a question on CPU architecture after I spent my time studying a bunch of C/++ trivia, I'm not going to answer the former as confidently since I'm pulling out knoweldge from a 3 year memory bank that I happened to work with at work.
>all the while YOU are the one benefitting from the system that cannot be changed.
I am? I'm a year out of a job and I sure wish I had a referral. My last studio shut down so those references scattered all over.
But for "whatever shall we do"... I don't have some magic perfect interview, but a few basic steps to save everyone's time:
1. we don't need 5-10+ rounds of interviews and 3 months just to make sure a candidate isn't a lemon. IMO if it's over 3 rounds (not including an HR call for basic alignment), there's some process that really isn't necessary. You dont need to speak to a CEO unless you're applying as a director or executive yourself. You don't need to meet every team member separately for their variation of an interview
2. Ghost jobs absolutely need to be regulated. I'm surprised it's even legal to lie to your shareholders and say "we're always growing" when half your postings have little intention of attracting a candidate. At the very least a job posting should delineate between urgent (hired withing 1-2 weeks of response), "normal" (2-4 week process), and "cold" (6+ week process, non-necessary role).
3. jobs should in fact give a general direction of what kinds of technical questions you expect to be asked for that portion. Who does it really benefit if one candidate happened to study leetcode (and the right kind of leetcode), but the other was focusing on system design, or questions about their specific industry problems solved? Or domain specific probblems?
Just a few ways to be honest, efficient, and bring out the best in candidates. Interviews shouldn't be so hostile to the labor the company clearly requires.
> if your questions can be memorized and recited by rote memory and the candidate can do well on it without the proctor knowing, then your questions are BAD or your proctor is a moron.
I agree in spirit. My only caveat sympathizing with interviewers is that most questions can indeed be "memorized", and arguably all questions/practices can be studied for. And there is some merit to trying to figure out how someone approaches a new problem. I think that approach is highly inefficient when you do in fact mostly want someone who "already memorized" the right domain knowledge, but I digress.
But people and their experiences are diverse. Without knowing their background (and again, some people seem to read the resume for the first time during the interview) you are never going to surprise 100% of your candidates with the same question bank. But that may be too much of a burden on trying to cater every inerview for every promising candidaet.
> Is the culture just so broken at these companies that it’s hopeless to expect people NOT to blatantly exploit the system for their friends? Why don’t people get fired for doing that?
How do you catch them? Unless they’re idiots they’ll write facially plausible explanations of their hire/no-hire decisions.
> the fact that there is a standardized test in my mind does the opposite and makes the process much fairer
The problem isn't standardized tests, it's that leetcode questions are about having the time to have learnt the answer beforehand, rather than raw ability for problem solving.
That’s not true. In any discipline when confronted with a test there are two strategies: brute memorization of the question/answers, or developing the skills to tackle the problem dynamically. You cannot categorically claim that LC tests are largely memorization tests rather than raw problem-solving skills. That is just the approach you are capable of taking. Not being able to see up the mountain doesn’t imply there are no climbers above you.
LC tests typically copy problems from the university the interviewer graduated from. College programs differ, so this is really a case of what you were introduced to.
There's a fairly popular online LC test company in my corner of the world which was formed by graduates and lecturers from a certain university and they started out by just giving the problems from the curriculum. Result was heavy bias in favour of students and alumni of that university.
>If that were the case then a normal, well accomplished software engineer shouldn't need to "grind" leetcode to pass an interview.
No, what this shows is that the skill range for accomplished professional software devs is absolutely massive. What these companies want is to find the tail end of this very wide distribution. Leetcode interviews do a decent job at this. If you have been coding for a decade and can't do leetcode mediums with almost no prep, and hards with a moderate refresher on data structures, then you're simply not the in the right tail of the skill distribution and they don't want you. This is what so many in our industry can't accept: you're just not talented enough to earn a FAANG job.
No, these engineers at FAANG companies could not solve those problems cold without having been taught how. I have worked at two, which is how I know. I have never seen a question in an interview I haven’t seen before. Many of these questions went unsolved for decades in the industry, so no, these engineers, who mostly aren’t DSA experts but distributed computing experts, could not solve them cold. I also saw how interviewers used questions to re-enforce their own biases on university, gender, or home country in these interviews.
I'm in a FAANG type company now, a YC company, 3000+ engineers. I'm a Staff SWE with 20+ years of experience (ECE degree) and make $600k+ per year. I've went through the promo cycle here (it sucks).
I can't do any leet hards and can do leet mediums after studying. Some easy's take me a couple of tries. I usually do very poorly in interview coding exercises.
Curious, would you say FAANG offers the right challenges to stay in the 0.1% or 1% if one started out there? Are they actually in the right place to grow?
depends. If your in at the bottom floor, then you will grow with the company.
If you are lucky and join/lead/get a new project off the ground, then you'll also grow.
If you're just trying to move the dial on a normal project, then its very difficult to make any kind of headway. You are surrounded by overly enthusiastic hall monitor types who will put in more hours than you, or post more than you.
I mean it doesn't, because I'm at a FAANG, At a FAANG you are infantalised from the very start, sure you passed a very difficult interview where you have to balance a binary tree efficiently as possible. But you're going to use none of those skills here.
what you actually end up doing is copy/pasting some random code you found using internal code search, because the sensible way of doing it can't happen as that would involve porting a thirdparty library, and doing all the procedural work that follows.
so you hack some shit together, ship out out and hope that it doesn't break. You then decommission the nasty hack you shipped last year and claim credit for innovating. Is your product not hitting the right metrics? loosing users? doesn't matter, so long as the super boss is happy that you've hammered in the REST API for the stupid AI interface, you're not going to get fired.
In a startup/small company, if you fuck up, the whole place is going under. Need metrics? you'll need to find a small, cheap and effective system.
Here, we just record then entire world and then throw hundreds of thousands of machines at it to make a vaguly SQL interface. Don't worry about normalising your data, or packing it efficiently just make a 72 column table, and fill it full of random JSON shit. Need to alter a metrics? just add a new column.
In short, don't praise or assume that FAANGs are any good at anything other than making money. They are effectively a high budget marvel movie, Sure they have a big set, but most of it is held together with tape and labour. Look round the side and you'll see its all wood, glue and gaffer tape.
>But you're going to use none of those skills here.
FAANGs want the top .1% of developers, they don't necessarily need them for most roles. But the point is to hire developers that you could put into any role in the company within reason and have them be successful. 99% of development work at a FAANG is pretty unexceptional and doesn't require exceptional developers. They hire for that exceptional 1%.
FAANGS want a load of loyal, naive people who are willing to work loads of overtime and not ask too many questions. Who better than posh kids from great universities who haven't quite figured out that life isn't a meritocracy yet!?
Sure they also want the top 0.1%, but they have a different interview track. Do you think all those OpenAI engineers that were going to follow Altman were asked to do leetcode?
Designing a test for which one cannot possibly prepare is a problem that’s bedeviled test makers for a long time. The team behind the SAT threw up their hands and said it no longer stood for “scholastic aptitude test” but just… nothing.
> You cannot categorically claim that LC tests are largely memorization tests rather than raw problem-solving skills.
Sure I can. By the time you get to Leetcode hard, these aren't just "can you derive the answer". The questions by design take 45+ minutes and have some weird quirk in it that is nominally related to the core concept being tested. These aren't necessarily meant to be done on the fly during an interview period.
>Not being able to see up the mountain doesn’t imply there are no climbers above you.
a better analogy is that youre on a road and you see a freeway above you. The people above you aren't "better", they are simply on another road, to another destination. But they aren't necessarily worse either. They could be on their way to a dead end job or could be a billionaire CEO.
That is to say, it's useless comparing yourself to other people you don't know. Everyone has their story.
Thank you. You just proved my point that “categorically LC is not largely memorization” by reinforcing that only in specific cases in some specific levels that you do need some specific domain knowledge.
My point is that LC hards were not intended nor designed to be perfromed in an interview setting, and essentially most people will only do the in that typical interview timeline if they've lucky enough to have seen it before or lucky enough to understand the specific domain knowledge.
So I'm not sure we're interpreting the same conclusion here unless you think trick questions are a mark of a "good engineer".
My point is that a claim cannot be made that categorically LC tests are generally a test on memorization. It appears your point does not disprove that. You mention exceptions, that in fact proves the general rule is true (I.e your claim
isn’t of the form: most leet code questions of almost any level cannot answered without spending 45 mins and requiring specialized domain knowledge as a prerequisite). You cannot argue against general rule by pointing out specialized exceptions. Cats generally have 4 legs as a rule. Yes exceptional conditions exists where they don’t have 4 legs, but the generally rule holds true. To argue against, your point must take the form: “as a rule, most cats do not have 4 legs.” Do you understand?
You can disagree, but I and many others can very much make a claim. I argue it can even be a well supported claim. Your personal disagreement isn't a refutal of any and all claims.
>You cannot argue against general rule by pointing out specialized exceptions.
Lertcode hards in interviews have been getting more common for a decade now. At what point do we stop pretending that more cats aren't losing their 4th leg and announce an epidemic?
It shouldn't be common, but it's becoming more common. I haven't seen any claims in any conversion even disagree with the notion, let alone provide hard evidence.
Again, my initial purpose of responding to the OP was his perspective is formed by his own ability, a trap you also fall into. Again, Higher mountain climbers and all. Lower climbers are more numerous and will agree and validate their experience amongst themselves as a group - and indeed they are the larger voice. Feel free to have the last word.
> The problem isn't standardized tests, it's that leetcode questions are about having the time to have learnt the answer beforehand, rather than raw ability for problem solving.
This sounds like you want to penalize students who studied for the exams. Or at least not reward them.
Like all interview formats, it’s a proxy for understanding if the prospect would be able to get the work done and be a good fit with the rest of the team. I’d say it’s a pretty good proxy for work ethic at crunch time as well.
If your complaint is that a normal person wouldn’t have the time to study these things in detail, why would a company want to hire someone who has external obligations?
> If your complaint is that a normal person wouldn’t have the time to study these things in detail, why would a company want to hire someone who has external obligations?
External obligations like full time employment and a family?
>This sounds like you want to penalize students who studied for the exams. Or at least not reward them.
Interviews are like exams but you don't have any clue what topics are on the test. If Leetcode was some licensed, standardized approach to getting some license to verify my ability to code myself out of a paper bag: I'd hate it, but I'd grit my teeth and study it. exams can be studied for.
But it isn't, so I can be studying leetcode questions and be hit by a dozen other topics. I don't have time to study everything, and the market right now isn't worth pinpointing specific companies unless you have a stellar reference.
>I’d say it’s a pretty good proxy for work ethic at crunch time as well.
I couldn't disagree more. Someone so prepared for interviews have the least skin in the game. Layoffs come and they didn't work > 40 hours but were otherwise excellent? oh well, get another job in a month because interviews are a breeze for them because they breathe DS&A.
>why would a company want to hire someone who has external obligations?
I'd love to know the answer here as well. Why do companies internally penalize workers who were laid off, but then try to "steal" currently happy employees but make them jump through these hoops? The logic seems backwards; interviews should be hard and depress wages for the laid off, desperate workers so you can get a desperate unicorn. But you want someone who lacks the time to study because of current job obligations to get to an offer faster. Their proof of work ethic is being employed to begin with.
The only value in leetcode is you should be able to solve a couple in a short time and thus prove you are least know something about writing code. We use them as an interview prescreen because once in a while someone seems like a good person we would like to work with, but we have no clue after the interview if they can really code.
We had one person who worked on [censored] 20 years ago, then was manager of [non-programmers, rest censored] - now wants to get back into coding - can this person still code? If so I want them, but if they have forgotten everything... I of course censored details for privacy reasons.
Because it seems a fairer way of apportioning limited opportunities than just looking at what school you went to, looking at previous companies on your resume, or seeing how well you can shoot the breeze.
> the fact that there is a standardized test in my mind does the opposite and makes the process much fairer.
It can make the process fairer but it's not a given. You can do the classic "no dogs, no blacks no Irish," and that was a standard across pubs in England. It certainly wasn't fair, unless you were racist.
If you're committed to fairness, then a standard will help. It gives you a clear point to fix and improve things and something you can use to measure if you're achieving your stated goals. That's definitely something you can't do if you're just making things up as you go along.
And yes, I made a deliberately provocative statement. I'm obviously not saying whiteboard tests are the equivalent of segregation.
A leet-code test would be much more standardized if candidates could solve it at home. Just send me a link to the quiz and let me solve it within a specified time frame.
I've done tests like this for some companies. It felt a lot fairer and more closely resembling the actual work environment than live leet-code interviews, with biased interviewer(s) and a stress factor that's not a part of the actual job.
As a hiring manager I HATE leet-code tests, and they do nothing to differentiate candidates, but a take home in the era where people run chatGPT beside the interview window, or have someone else do the interview for them? Not a chance. You are 100% correct that it is way more representative, but the prevalence of cheating is ridiculous.
I totally understand you, but want to offer a different perspective.
They will also be able to use ChatGPT on the job. And StackOverflow. And Google. If they know how to use tools available to solve a problem, that will benefit them on the job.
If you're testing them for what ChatGPT can already solve, then are the skills being tested worth anything, in this day and age?
Take-home LeetCode, even with cheating will still filter out a good chunk of candidates. Those who are not motivated enough or those who don't even know how to use the available help. You'll still be able to rank those who solved the task. You'll still see the produced code and be able to judge it.
Like other commenter points out, you can always follow up the take-home LeetCode. Usually, it becomes apparent really quickly if a candidate solved it on their own.
This does seem like a vexing problem, especially when interviews are conducted remotely.
I wonder if either of the following could be cost-effective:
(a) Fly the candidate to a company office, where their compute usage could be casually monitored by an employee.
(b) Use high-quality proctoring services that are nearby to the candidate. E.g., give them 1-2 days in a coworking space, and hire a proctor to verify that thy're not egregiously using tools like ChatGPT.
Or alternatively, would it suffice to just have a long conversation with the candidate about their solution? E.g. what design trade-offs they considered, or how might they adapt their solution to certain changes in the requirements.
> I'm sure anyone determined to do so can act unfairly regardless of what process is in place, but the fact that there is a standardized test in my mind does the opposite and makes the process much fairer.
This is what privilege looks like. The inability to see barriers that affect others in a worse position.
Standards do not imply fairness only consistency.
You got a test that filters out people who lack the time to train for these tests. Basically, devs with a life (see family)
OK, but I have a two-year-old, so consider my privilege sufficiently "checked" before having written the post. Anyway, what would you suggest that would be more fair?
Within one organization the form is standardized, and across companies it's very similar. But sure, maybe not the right word. I'm comparing it to "just letting the interviewers chat and decide based on whatever" which is typical in industries without a skills test.
Leetcode-style tech interviews fall far short of standardized tests. In real standardized tests, the same questions are given to everyone, and effort is made to develop new questions not known to the test takers.
They're more like a standardized test than a pure shoot-the-shit interview is, but still pretty far away.
yeah i agree that "if you know how to problem solve you will pass" statement is a joke. you absolutely need to memorize most of these problems as you'll never encounter them out in the real world. i think we need to get better at the behavioral side of interviewing - this should be the juice of getting at whether or not an engineer is good. and if they're really good at lying... CEO material? lol.
The problem you are describing is interview variance and hiring bias, not leetcode. This happens irrespective of interview style.
Many companies have question banks that are specially designed to be fair/have some contextual relevance (ideally) to some "realistic" problem. Or at least, many of the companies I've interviewed at follow this model. I consider these coding questions to be "leetcode" style because at the end of the day they are an isolated coding problem, even though they may not be a problem from leetcode verbatim.
Companies that execute on that style of interview well are generally fairly pleasant interviews, at least in my experience. Good companies/interviewers will gauge more than just the final code to determine a hire or no hire. And a large portion of companies also have hiring ratings on a scale to make it less binary.
Question banks that are too big: huge variance, and OP's point stands.
Question banks that are too small: leaked on eastern forums immediately, candidates show up reading answers out to you (some of the guides include guidance on when to pretend to think, I am not kidding).
The idealized version of "question banks" might work. The real one does not; you'd require employees constantly scouring forums in every language known to mankind, immediately removing anything that gets leaked. On top of that you'd probably require a competent committee overseeing all questions in the bank constantly and ensuring the lack of variance in difficulty.
I agree with you, but I don't really see how this invalidates the style of interviews where you're presented with some timeboxed coding problem (of reasonably scaled difficulty) and are asked to solve it.
There will be bad actors regardless of the interview style, thats why companies have multiple interview types/styles/rounds to sus out a candidate, as you probably know.
If they BSed their way through a leetcode interview, then they probably won't make it past a behavioral interview where they have to go in depth on some past project. And if they BSed that as well as every other round, then hey maybe they are crafty enough to succeed at the actual job.
I think this is where our different opinions come from, while we agree on the other aspects.
In my personal experience, I have never felt that the hire/no-hire decision relied exclusively on my ability of solving the presented problem; I have passed interviews where I did not solve the LC-style problem optimally but I communicated clearly, picked up on hints, was aware of when I hit "walls" and provided working but less than ideal alternatives when I could not figure out the neat tricks.
Reading through the thread it seems that my experience is not universal, and the majority here have had less pleasant interviews, so I understand where you are coming from.
I have had all possible experiences. Sometimes I feel like genius and ace some leetcode with an almost novel solution. Sometimes I missunderstand the question/scope and mud myself into the hole of despair.
I have been rejected for one mediocre interview among many good ones. Or the other way around accepted even though I didn't perform well.
Sometimes the interviewer works with me. Sometimes against me. Sometimes a war story impresseses positively, sometimes raises suspicions.
At this point it feels like gambling.
I have also ran almost 400 interviews on the other side over the years, and to me it seems quite clear when somebody cannot write code at all. I like to think I am not biased. But who knows.
>Reading through the thread it seems that my experience is not universal, and the majority here have had less pleasant interviews, so I understand where you are coming from.
it changes immensely based on the job market. I've defintiely tanked some inerviews hard, stumbling on softball questions that shoulda been a bullet point. But I get pretty far or even gotten offers.
The last 12-18 months though? I've had interviews that felt like a dream but got zero follow up to. Been ghosted after seemingly final rounds where I spent 5+ hours on technical tests. It's not even enough to "understand the problem and communicate steps". You gotta be flawless, and you still might be cut compared to 3 years ago where a "C" performance could still land multiple offers as long as your experience made up for the quiz questions.
I dodged the .com bust because I worked for the U.S. DoD at the time.
But I got laid off for the first time during last year's "15% bloodbath".
If I compare my current job search vs. all of my job searches in the past:
(1) As parent comment said, the bar seems to be much higher. I've thought that I did really well on some interviews, only to not get an offer.
(2) Some interview processes are way more rigorous. For a DevTech role within nVidia, I had 12 interviews + 2 take-home problems. (BTW, the take home problems were incredibly fun. Well done nVidia!)
(3) I've finally accepted a job offer from a large, established tech company, and the pre-onboarding process is amazingly slow. I accepted the offer a month ago and still don't have a start date. In a better job market either (a) they'd probably work harder to be good about this stuff, or (b) I'd just take a different job because of the delay.
> (2) Some interview processes are way more rigorous. For a DevTech role within nVidia, I had 12 interviews + 2 take-home problems. (BTW, the take home problems were incredibly fun. Well done nVidia!)
That's ridiculous, tho.
Did they really not have enough information on the 11th interview to know whether or not they wanted to hire you?
Out of curiosity, can you share some examples from eastern Europe? Where does one find such websites?
Also, IMHO, blanket generalizations like "Asia" and "Eastern Europe" in such contexts can actually be more offensive than just mentioning the one country where the thing happens since you're basically painting with tar a whole sub-continent with dozens of different countries, just for the things happening in one country.
What I mean is, if by Eastern Europe, you actually mean some dodgy Russian forums, I think a lot of Eastern Europeans from Bulgaria all the way to Poland and the Baltics might feel offended of being included since we are not the same.
No, because I feel like mentioning previous employers AND mentioning the languages I speak would get quite specific.
You are interested in Eastern Europe, so you should be familiar with olympiads; ask around any circle of ex-olympiad participants and you are bound to find something.
I am not sorry if you took offense; you're either from one of these countries and are clueless to the circles that exist next to you, or you are not from one of these countries and are trying to be offended on behalf of others.
Yeah, there's also the implication here that Americans rarely cheat. They aren't as public because English is under a microscope, but there are definitely answer banks if you know the right person and can fork over the cash.
If it's anything like High School/College, the sad part is these kinds of people could probably do well in interviews regardless. These answer banks are simply the difference between an A and an A+. And sadly the current market seems to only want A+ candidates. Who's fault is it really?
> And sadly the current market seems to only want A+ candidates.
You mean the current FAANG market paying top dollar. I know plenty of unknown companies taking B candidates because they aren't paying top dollar (in Europe at least)
>Who's fault is it really?
The governments and central banks for devaluing the currency post-2008 with their zero interest rates, causing the value of savings and wages to plummet and the value of assets, housing and stonks to skyrocket, causing people to chase get rich quick schemes on the stock market and on the jobs market. Coupled with the VC promoting for an unsustainable growth (more like pump and dump) of so called "tech companies" who's products were not economically viable from the get go, they just survived on zero-interest money and fake promises, artificially boosting the demand for SW workers casing many young people to go into tech just to chase money, money that's now gone and so is the demand for coders.
Without this artificial demand for devs caused by zero rates and overhype in the tech industry of financially unsustainable products that were banking on skirting the local laws (AirBnB, Uber, etc), then those people chasing money would have went into finance or investment banking to chase money there instead of causing a huge backlog of candidates in tech. Just my 0.02$
> I know plenty of unknown companies taking C-/B+ candidates because they aren't paying top dollar.
in my experience: not these days. And I work in games, so "top dollar" was never in question.
>My 0.02$
That's a fair take. Tech in some ways was indeed a necessity with a huge reach, so I don't think the overhype was the difference between being a trillion dollar industry and a million dollar one. But tech would probably still be a billion dollar industry without all the factors you mentioned.
Games... well, stability was never a factor, and I knew that going in. It's a shame they are doing the exact same pump and dump schemes tech has fallen into. And it's not like layoffs were uncommon after a project finished; it's just that they are doing it purely for better looking earnings call instead of "we cannot keep funding studios anymore".
So New York, Boston, Washington, Baltimore, Philadelphia, Washington DC, Maryland, Virginia, Delaware, New Jersey, Main, Connecticut, Rhode Island, etc?
Or by "cheating" are you specifically referring to the lying cheating treasonous fraud from New York who was just found guilty of 34 felonies to cover up cheating on his wife with a porn star to inflence the election?
Washington is a state on the west coast. Washington D.C. straddles Maryland and Virginia. Baltimore is a city in Maryland.
Why are you double-dipping on examples? If people don’t know their geography it’s meaningless, and if they do (like me) it makes you seem disingenuous.
Oh but then you bring up politics so it makes sense. Fuck this bullshit world, it’s so exhausting.
I take issue with it because Leetcode gives tech people a fake feel good, “yeah but at least it’s fair” illusion, when really it’s probably just as biased as any other hiring method.
This is ironic considering these companies have forced mandatory DEI seminars (which I have no problem with btw), inclusive language, #EveryoneCanCode, and so on.
But despite all this, you end up with teams and organizations that are 99% of the same X somehow. Replace X with race, school, even state from home country sometimes.
You know there are websites where people share all the interview questions to these hard interviews / referrals exclusively in their language and behind a pay or rep wall? And there are Telegram groups where international students leaks the questions or do interviews in place for one another.
It’s inevitable these types of issues arise when there’s so much at stake, ex: in just 5 years, a $200k TC advantage at a top company, becomes $1m or more with appreciation.
I just really dislike the veneer of “fairness” when there are so many problems with the process, even beyond the questions that have nothing to do with the job.
> This is ironic considering these companies have forced mandatory DEI seminars (which I have no problem with btw), inclusive language, #EveryoneCanCode, and so on.
I really do wonder what those sanctimonious sermons are meant to accomplish. People who are already ideologically aligned with them won't learn anything new and may just resent it, while people who aren't aligned won't become aligned as a result of that "training".
But you're talking about it as though you expect it to have a constructive impact, resulting in irony when it doesn't. I don't see the irony because I don't expect any benefit from those struggle sessions in the first place.
They're supposed to absolve the corporation and its leadership of responsibility and liability. By giving that training they get to claim they did all they could.
Struggle sessions in their basic form used social coercement to extract confessions of guilt against some collective cause. This describes the DEI training sessions I've been in well. Admit that you have bias (effectively confessing guilt to a "crime" that gives your employer leverage over you) or get dogpiled and have your refusal to admit guilt cast as evidence of your guilt anyway.
You'd be surprised how much self-perception and reality can differ. Lots of types that think they're the best allies ever and then show a total lack of empathy (especially if someone's different in their immediate family) that'd need a reality check.
Though usually the intersection of people running DEI initiatives and those people make a large set. Assimilated gay people love talking shit about how trans people make "us" look bad.
Engineers typically optimize for their own learning, and the false negative rate can reflect more about how aggressively the interviewing panel wants specific peers versus how much they actually want the job done well.
Suppose there was a SWE union and the union has to create their own reference compensation bands. Would the union again lean towards leetcode or some other standardized test? Or just use years of experience and maybe past projects?
When weighing the “effectiveness” of leetcode interviews, it’s important to weigh that thus far SWEs have failed to effectively unionize despite e.g. the past class-action no-poaches clearly showing C-suites should be paying engineers a larger share.
I don't think the problem is the format, i.e. a 30-50 minute interview on simple coding with DS&A problems, but the escalation.
The reality is, fizz buzz got us 75% of the way there. It turns out when pressed, a lot of people can't write code. Yes, there's false positives, but there's also people brute focing their way through via copy & paste.
This doesn't manifest as a person who can't do any task, just as a person that's slow, delivers weird abstractions, and would take a lot of your time to get anything useful from.
But those people are also making those arguments, because, as you said, there's hundreds of thousands of dollars in it.
Yeah I know how to implement FizzBuzz since it's such a meme, but I've basically never used the mod operator in real code. Maybe it comes up in more math-y code I suppose, but for most backend/frontend/SQL code I've never reached for it.
I’ve used it for coloring alternating lines differently in UI code, and as a lazy way to log only every so many times some loop runs.
I only know it well because it was covered near the beginning of one of the first programming books I picked up (on Perl 5) and it stuck with me because it seemed wild to me that they had an operator for that.
Even if that's so, modulus (or at least the concept of remainders) are elementary school math and any competent programmer could bang together an (inefficient) modulus operator in a few minutes.
So even in a language w/o a mod operator, it's not a hard problem if you understand how to solve problems with code.
You don't actually need mod to do fizzbuzz, even if that's the most obvious way for people who know what mod is.
But without any "real" math at all you can do it with, eg, two counters and some if statements. Or if you recognise that there's a repeating pattern you can work out that pattern manually and just write code to emit it over and over.
I once worked with a guy who was an incredibly good developer and I was surprised when he didn't see anything special about the number 64 (i.e. a power of two) - turns out that he'd never done any bit fiddling type work so he hadn't had to think in those terms. It wouldn't surprise me if a lot of people hadn't heard of "mod" either....
A huge majority of programming work is basically just CRUD stuff and other data shuffling. It’s not surprising that someone wouldn’t have needed to work with big shifting (or modulus) in that case.
He was an expert in complex multi-organisation enterprise integration and was the go to guy to work out why horrific distributed transactions were failing... He also did a lot of cool stuff as side projects in his own time - just none of them happened to involve worrying about powers of 2.
Unless you specifically want a compiling and running version of FizzBuzz you don't actually need to use or know about the mod operator.
At least for me it would be sufficient if the person used a function like IsMultipleOf(x, m), or Remainder(x, n). This would at least make it clear what the function did even if they didn't get the exact operator.
The other thing to note is that the mod operator works differently on different languages and platforms.
> the mod operator works differently on different languages and platforms
Not with positive inputs, which is the domain of FizzBuzz.
Even if you don't know what "mod" means, if you have no idea know what a remainder is, and that the problem calls for it, and you can't derive the mod operator using integer addition, subtraction, multiplication, and division, then your math and problem solving skills are pretty weak, which FizzBuzz tests.
I stopped using fizz buzz a long time ago. 90% of candidates can't define a 2d array in their chosen language without first filtering the candidates, where you get to about 50%.
Same experience, used FizzBuzz at many places and always got surprised by the amount of people it can filter. The best interview process I've ever ran at a company consisted of a basic FizzBuzz for about 15-30 min followed by a pairing session no longer than 1h30m on a problem that could get as tricky as we wanted to assess their skill level.
We would both test the basics as well as go through with the candidate on how they think, how they collaborate, help them out if we felt nervousness was impacting them showing their skills, and in the end got a much better grasp on how skilled they were than if we were looking at Github repos or giving DS&A trivias to solve.
I once froze in an interview when asked a simple technical question - I'd been giving a presentation for an hour on how to launch a new product and I was asked by the CEO how to do something technically trivial - my brain could not do it. So he probably thought I was some marketeer pretending to be technical - which isn't really true.
I suspect quite a lot of people who are labelled as "can't code" are freezing like I did.
- give them ~30 minutes on their own machine
- they get to pick the language they code in
- I leave the room for large parts of it
- I bring them a drink
- I tell them I want to see what they can do and that I don't care if they complete it
- permit them to search on the internet (as long as they don't copy/paste a solution)
I see this usually:
- they finish in a few minutes
- they just can't do it, even with hints
I suspect some do, and some will be a lot when you bunch them all together.
However, counter point, I've had people forced on me through much of my career who just can't code for the most part, and despite being pretty reasonable about it ( it's part of the nature of the work I do ), I'm very rarely surprised by competency.
I failed FizzBuzz the first time someone gave it to me in an interview...
The specific failure was that I first attempted to solve by using repeated subtraction. The guy kept asking me to "solve it a different way", or saying "there is a better way to solve this". I tried using arithmetic tables, I tried using results about base10 remainders and I even tried using one of the corollaries to Fermat's little theorem to speed it up for larger inputs... every time I was told I was getting it wrong because "there was a better solution".
In the end he pointed out that the only solution he would accept was use of the mod operator.
Since then I have actively kept a tally: I naturally use the mod operator an average of twice a calendar year, it has always been in personal life code when dealing with complicated geometry problems, the bit of code containing it almost always fails on some edgecase because at the point of using mod it is convoluted.
Honestly sounds like a bad interviewer. repeated subtraction is a good first step and I would try to push more if that was the first implementation. But If you could derive a base 10 remainder you know conceptually what problem the mod operator is trying to solve.
a % b = a - (b * a/b) /assuming a sane language with integer division, else cast a/b to int/
Figuring the above operation (or getting close) is when you should more or less pass, and That's a good point to show the interviewee what the operation is. That should be the point of an interview problem, to show the thought process, not that you know fancy tricks.
But alas, I was shown an XOR swap in an interview last week and spent 3 minutes deriving it on paper instead of 3 seconds saying "oh yeah, a => b and b => a" to a trick that I haven't seen since college some decade ago. The current market loves tricksters, I suppose.
And yes, the actual real world use of modulo is surprisingly sparse, despite easily imagining situations where you could use it.
I am highly reluctant to use type casting as a mathematical function!! I was burnt by early languages struggling with this problem... Even modern languages still have issues with this! Try running this in Python3.11
FizzBuzz is a highly artificial problem. It makes sense that people who are not familiar with it will assume that there is an elegant solution. But in the end the right approach is to be very boring and to notice that you need to check for divisibility with 15 before you check for divisibility with 3 and 5.
FizzBuzz is a problem that doesn't have an elegant solution. That is the point: to see how you approach the problem. (there are 3 possible solutions, each in-elegant in their own way)
I don't like FizzBuzz because it over-weights the interviewees knowledge of the relatively obscure modulo operator. Yes, there are other ways to do it, but the expectation of FizzBuzz is that the candidate has that "Eureka" moment and remembers modulo.
If I need a "Non-Programmer Weed Out" question, I'd rather give a problem that is 1. as easy as FizzBuzz, but 2. is just 'if' statements and loops and cannot be solved by knowledge of a particular operator (or bit twiddling tricks).
A lot of people couldn’t even read this thread if someone was watching closely. They’d recognize words and idioms, but couldn’t make sense of these or think. It’s called anxiety and rarely has anything to do with actual performance. Anxiety is a frequent guest in a smart guy’s home.
As an early 2000s integrator/analytics I learned to write code when someone watches out of interest or straight time pressure (still taxing sometimes). But most developers that I know intellectually curl up in such situation, regardless of their skill or performance levels. We worked with one very math-y low-level-skilled guy whom our clients described as “literally freezes for an hour without moving, should we pay for it?” when he did field work. He was a very strong developer otherwise.
Not that a company must hire or want these people, but the idea of writing code under uncanny pressure all day makes as much sense as that swordfish scene.
I have empathy for the false negatives. I have been them, or maybe at times I'm simply a false positive, the problem is those people are likely to freeze no matter the interview.
Further still, I get push back with folks citing self-prescribed medical conditions, but the same people generally display the same behaviours during the working day.
So other than contract to hire, which typically limits you to people with a job, I don't personally have a better way.
> I have empathy for the false negatives. I have been them, or maybe at times I'm simply a false positive, the problem is those people are likely to freeze no matter the interview.
I can't speak to the statistical claim that "those people are more likely to freeze no matter the interview."
But just my personal anecdote: I do fine on take-home coding tests, but I freeze up on live-coding tests.
This is one reason I'm vexed by the allegedly common cheating in take-home coding tests. It makes employers suspicious of the testing style that I'm best at.
I've actually had people cheat on live coding sessions.
For north of $2m over your career, cheating probably is the smart thing to do, especially for a borderline candidate, as there's a fair amount of evidence that the prestige on your CV will make things easier going forward.
However, my problem with take homes was never that the candidate would cheat, but rather they'll probably spend way more time than the 2 hours allocated.
I'm actually less worried about the candidate doing that, than I'm worried that the interviewer bakes in a bunch of assumptions like having a machine setup to do the task, having the specific domain knowledge and experience, and then accidentally trolls the candidate with little to no avenue for feedback.
I totally agree with A and B and have further suspected they use leetcode as a means to support a H1B candidates employment prospects (for those who do pass)
There is no grand conspiracy. People try to develop what they feel is a fair interview with good signals. It simply turns out that many people are not good at that and have biases that they may not even be aware of. Tackle on some cargo culting and "best practices" and you have your typical broken process.
For those wondering what the alternative is to leetcode: a work sample test. Take a real problem your team solved recently, distill it down to remove internal-specific knowledge, and give candidates 3x the time to solve it compared to what it takes an internal dev to solve. Bonus points for having a rubric guide the scoring.
I don't buy the analysis on 1). Leetcode interviews restrict the pool of SWEs. You have a lot of SWEs who fail to qualify any more, but the few that do are more coveted and get to command higher wages
What about a type of blind interview where neither of the two know the problem or solution in advance. After, their solution is blindly reviewed/graded by a third party.
- Input from the interviewer reflects communication ability, cultural fit, and so on of the candidate.
- Input from the blind grader reflects ability of them as a team.
- Low grades count against the interviewer as well.
This is very much like the "real world" situation. OTOH, I think this leaves the interviewee even more vulnerable to the unfortunate situation where they correctly solve the problem but the interviewer is convinced that a different, wrong solution is correct. Live long enough and you'll have that happen to you.
OTOH, seeing how the interviewer reacts to that situation will tell you a lot about whether you really want to work there.
> To the company this is acceptable noise, but to the individual, this is costing us 100s of thousands of dollars, because there’s only a handful of companies that pay well and they all have the same interview process.
If you want to earn top money, you gotta play the game. This is true in every profession, you think lawyers that want to work at top law firms have it any easier?
There's a lot of valid criticism with respect to today's tech hiring practices (especially because it doesn't only affect top paying companies), but I don't feel like "I can't get ridiculously wealthy without selling my soul" is a particularly compelling one.
Although I find it disappointing, this is the truth. It is a function of supply and demand. There is always a gating function when you have an abundance of supply; sometimes it is leetcode, sometimes it is school, take home work, resume formatting, a mixture, all of the above, etc.
Everything is a proxy because, in most scenarios, there is no way to understand effectiveness of an employee until they have worked for you for a while.
Of course there are other ways, but they all carry their own risks.
That’s why referrals, imperfect as they are, carry so much sway. In theory they reduce uncertainty and require some expenditure of political capital.
Naive questions ahead, if hires weren’t made scarce by some absurd filter, why would they pay $200-300k extra? Feels like the whole idea of stellar salaries must be based on something stellar. Afair, before Google(?) made it normal, developers were sort of dirt cheap. Weren’t developers in abundance at all times?
The reason they pay $200-300k extra is to attract the best they can. Say you got the same salary working in an ethical company than in a FAANG, would you go for the FAANG?
The absurd filter is just some kind of lottery. They could have a different one: at the end of the day, it's only when the person starts working that you can actually see what they are worth.
The thing with a filter like this one is that it filters out a whole category of people who may actually be good. And it reinforces itself: you hire people who are good at leetcode, who will themselves possibly be good at hiring people who are good at leetcode. Does a company of leetcoders perform better than a company made of a diversity of good engineers? Not clear.
I agree with this. But the root commenter talks about it as something these people are worth, when it’s just that - a lottery. A company that has so much money that they don’t know which barrier to put there to stop the flood. You jump through absurd hoops and join the club. Kinda like being accepted into a cool kids league, but activities are the same.
> Say you got the same salary working in an ethical company than in a FAANG, would you go for the FAANG?
Very much a tangent, but what do you mean by a company being "ethical"?
I have a few concerns about how that term is often used in these discussions:
(1) it's treated as a binary quality, rather than some kind of continuum, and
(2) there's rarely a recognition that different persons have different conceptions of what's ethical, let alone a justification for what the writer's preferred definition is uniquely superior.
The proof of some of this is that these companies could come up with a standardized test maybe with, say, smaller-scale refreshers every few years (this is normal in some professional certs) and save a ton of money if it were really about ensuring a base level of competence using these sorts of questions as the yardstick.
But that would dramatically increase worker mobility, for one thing. So they don’t do that.
(“No it’s because they don’t want you to be able to test prep” that makes no sense because 99% of the way to prep for these is… exactly like test prep)
This is a good assessment. 1 and 2 are why the system won't change, but I don't think they were intentionally designed with that in mind. I think it's a hang over from academia, bearing in mind how many of the top engineers at FAANG are PhDs.
Well, that and how almost nobody who successfully finds employment after "grinding" leetcodes wants to remove the barriers for entry.
I think Leetcoders can't envisage a better way to assess someone than by subjecting them to the same kind of hoop-jumping you get made to do in university. They're not interviewing you for a job as there's no module on interviewing candidates on the CS curriculum, and don't have much professional experience outside of academics or software engineering. They're simulating a dissertation defence, because that's how they were assessed for their competence.
That's my charitable interpretation. If I'm being cynical, it's elitism - a way of making sure you're "one of us" (read: obnoxiously academic, Type-A personality, "logic over feelings").
> how many of the top engineers at FAANG are PhDs.
Not that many, but its interesting how many are from "elite" universities. I'm in a research org for a FAANG and we often get to see all the handwringing about how we can't recruit more of people type x.
Well, if you only hire from MIT, Stanford, Oxford etc, then they are all going to look the same.
For those outside the research org, its a bit better, but its still the most uneven place I've worked.
I interviewed at a company known for consistently asking one of the same four questions in a specific interview round. These questions were widely shared on forums like Blind, Leetcode, and Glassdoor. The recruiters also provided strong guidance on the type of problems to expect.
I prepared thoroughly for all four main questions and any other plausible ones I could think of. I practiced writing solutions to ensure I was fast enough for the interview. Additionally, I pre-prepared ideal answers for each question in case I got stuck.
When the interview came, I got a total curveball: a question that was significantly harder than the usual ones. It didn't fit the round's theme (it was a DSA question, but I'd already aced the DSA round), was obscure enough not to be on LeetCode, and required writing a solver for a hard variant of a known algorithm. I panicked, copied the prompt into ChatGPT (despite being instructed not to use it), transcribed the result, and pretended I had recently studied the relevant algorithm.
I passed the round, nailed the other interviews, got the offer, and accepted. Later, I found out that interviewers are instructed to pick one of four specific questions for that round, and the one I got wasn't in the list.
I'm left wondering if the interviewer was trying to sink me or was just bored with the usual questions. The whole experience raised several questions for me:
Is it cheating if I already had pre-prepared answers for the questions they were supposed to ask? What's the difference between using pre-prepared answers and using Google or ChatGPT during the interview?
If the interview had gone according to plan, what was I actually demonstrating? My ability to use Google?
When the interviewer asked an impossibly difficult question, I would have failed if I answered it legit, even though I'm a good engineer. Failing such an unfair interview round doesn't serve the company's interests.
What is this interview process meant to demonstrate? My true value as an engineer lies in my ability to communicate clearly, think outside the box, identify and address technical tradeoffs, mentor juniors, and propose technical solutions that meet requirements while minimizing risks. Yet, I'm expected to solve a hard variant of the Traveling Salesman Problem in 45 minutes or I don't get the job? Why?
The whole process seems broken, but I'm not sure how to fix it.
This reads very similarly to consulting case interview questions. 1) It requires a lot of free time beforehand to study the frameworks so it filters intent but also removes those who need to work and 2) it seems objective on the surface but you can essentially pick winners by choosing difficulty and nudging along.
I'm not sure there's so much agency in any of this. Most of it is probably cargo-culting. Every software shop wants to be Google, so they do what Google does (or what they think Google is doing). It doesn't matter if they do it well, or understand (or agree with) the ultimate purpose of the process.
If Google engineers all wore pink robes and top hats while interviewing then we'd see that everywhere.
I think you’re overthinking it. Hanlon’s razor applies - the industry as a whole is incompetent at interviewing. And to be clear I’m not being smug here - every time I’ve ran recruitment rounds the interview process was a mess. It’s an extremely difficult equation to balance and as an industry we have never made the commitment to figure out a golden standard for how to do it.
>because there’s only a handful of companies that pay well
If this was just FAANG I wouldn't even mind it. Bigger compensation means bigger hoops to jump through, artificial or otherwise. At least those bigger companies come with interview guides.
But I've been dropped Leetcode mediums on some Unity game dev jobs barely paying 40/hr with minimal benefits. Game interviews already have an absurd number of topics to try and guess you'll be quizzed on. (last interview wanted low level C memory/bit manipulation question... the one before that did Unreal Engine trivia... then the one before that decided to give me vector math and ray intersection questions), adding in random string manipulation questions when you will barely use more than concatenation on the job doesn't help with studying.
If you are applying for game dev jobs, C memory management, the gnarly parts of Unreal engine, vector math, and ray intersection are all things you could be doing on the job.
Game dev is lower paying, less structured, and more involved than typical web dev jobs.
I’d really only want to get into game dev early in my career or as an independent creator. It’s a touch exploitative.
- graphics/shader programming ( I do enjoy graphics programming, but I have been asked these questions despite not applying for graphics roles)
- GPGPU
- traditional software engineering (programming patterns, architecture)
- netcode (I never applied to a network engineering position. I in fact actively avoid that part of the stack)
- general industry questions (what kinds of bugs can appear on shipping build that may not appear on debug builds?)
- a myriad of gameplay programming paradigms
- coding tests in Unity, despite applying for an unreal engine position (this has in fact happened twice now. I know Unity, but come on: can you imagine getting a javascript test for a c++ position?).
I've been asked all of these in some capacity over the years. At what point is it on me for not being some sort of omni net/graphics/engine gamedev that can answer any question on the fly, and not on the company for being paranoid about professionals studying for the skills they are clearly seeking? If someone can "study for your interview", why wouldn't you want that person? That's what they do on the job after all.
> I’d really only want to get into game dev early in my career or as an independent creator.
I'm already 8 years in, so a bit late for that. My plan is eventually to go indie, but I need a bit more time in my career before making that jump.
Despite the common advice c. 2022, getting a boring cushy tech job right now isn't very viable, so may as well stick to what I'm actually getting interviews for.
Fair enough man. I've done hobbyist game development, independent game development, and applied to some shops myself early on. But eventually, you get a job and your path diverges, etc, etc. Like, I'm not anywhere I'd imagine I'd be when I was in school.
I think at least half of all programmers got into it because of video games. They know it's desirable, so they can filter almost as hard as they want.
Like, the industry is just way more jacked up than general software development. I understand your pain.
Although, if you're currently between jobs right now, you could crank out a prototype of something and see where it takes you while job hunting. Because you will always feel like you could use just a bit more time. Sometimes you just gotta do it.
>I think at least half of all programmers got into it because of video games. They know it's desirable, so they can filter almost as hard as they want.
I had this whole pipeline where I figured
- okay, I don't have a master's degree
- there aren't any junior graphics roles.
- I'll work in industry for 4-5 years in whatever I can while working on rendering tech, and then get a graphics programmer job this way.
- I'll be in a cushy, in-demand, highly skilled, more protected part of industry from the layoff hustle and bustle. You don't throw away engine programmers at the blink of an eye
Alas, the industry shifted around in the blink of an eye. Most studios cared less about fundamental graphics for an in-house engine as they all just changed to unreal. I did in fact get an engine programmer role, but was still first on the chopping block for layoffs. The industry tighted the heck up over the last 18 months and my side hobby projects on github barely mattered as they wanted 5+ years of graphics experience. Back in the catch 22 I was trying to avoid with all this prep.
I wouldn't say my dreams are crashing down, but it sure did show me that nothing is truly safe, no matter how experience or well liked you are. Nor even how exploited; definitely had freinds completely underpaid that put 80+ hours into the company, and they still get cut the moment RTO is announced. You have the dream capitalist candidate but can't handle not seeing them 5 days a week with butt in seat.
>if you're currently between jobs right now, you could crank out a prototype of something and see where it takes you while job hunting.
That's the weird part. Prototypes are normally a great way to stand out. But the market here is sending all kinds of weird signals. I don't know what companies want anymore, clearly not a plain ol' hard worker.
I am still getting interviews, but they are cut short in the process or am simply ghosted. Some are botched tehnical interviews, others I have no clue. So more prototypes aren't necessarily solving my personal issues.
We’re hiring for an entry level position where I work. So entry level, that their first month is going to be dedicated to simply learning the framework we use.
However, we do require a base level of competency. We give out a 40-ish minute assessment. Two multiple choice and three coding problems.
And they’re easy. One of them is essentially “John has a 5 gallon bucket and 3 gallon bucket, how many buckets does he have?”. All you have to do is rewrite the clearly labeled circuit diagram as a Boolean expression.
Not a single person this round has passed. Several have failed the simple ones.
So while grilling people on the minutiae of a language or asking them to solve the traveling salesman is not beneficial, neither is nothing.
It's also a great way to filter out anyone who has any kind of commitment like kids, aging parents to take care of, or have health issues themselves that makes it hard to cram leet code non stop on top of a busy career.
For companies this is very rational, you want non-distracted employees who can work over time.
It's really not that rational unless you want someone only familiar with being the top code monkey.
In all my years of sw dev the number of times that would have helped vs being able to communicate and manage expectations across a swath of people is like 1:1000.
That statement is not untrue, but I think it over-estimates how much of the "actual work" is banging out code, vs. making sure the correct code is being written.
Yea, I've worked in a place where it was just "coders coding" and nobody was communicating and managing expectations, and that was its own unique form of awful. You need both.
> there’s only a handful of companies that pay well
with all due respect I believe this is about as far from the truth as it gets. I believe the issue is that people think they have to work at FAANGs in order to be paid REALLY well which is just nuts.
I know many people making (very) high 6 figures working at places most have never heard of. instead of looking for FAANGs people should be looking at companies that have been in business for 20+ years (if publicly-traded company the easiest "measure" would be whether you would invest your own dough into that company) and then make your self indispensable there (this is much easier than it sounds) to a point where you are more valuable to the company they company is to you. This is un-achievable at FAANG but this should be everyone's career goal. once you are worth more to a company then company is to you, compensation-wise sky is the limit :)
> I know many people making (very) high 6 figures working at places most have never heard of
You know people making ~800-900k at places most people have never heard of? That is what "high 6 figures" means. I have to assume you mean more like ~$180-190k?
Not op, but FWIW, I know people who make high six figures (explicitly, $750k>) at companies that aren't household names or FAANG(ish) companies (or AI companies for that matter), though I also know my share of people that are at such companies that most people, or maybe just people here have heard of, that also make > $750k. Some of them aren't even software engineers! Most (all?) of them aren't paid that much on their W2 though and have lucked out with stock options/RSUs/related compensation (again, at non-FAANG companies, some private).
Not trying to brag that I have well-paid friends, though it might come across that way; just corroborating the fact that it's possible for late-career professionals to make that kind of money at non-FAANG companies.
I 100% believe it's possible, but that's even more of a lottery than FAANG is which would negate his point :).
If you can get your foot in the door it's hard to beat FAANG comp on a risk-adjusted basis and they do actually employ quite a lot of SWEs, which I think is why it's such a focal point for tech jobs even though you can get similar comp in other situations.
If you start to view yourself as a business vs. an easily-replaceable-one-of-thousands individual your career takes an entirely different shape.
the entire system from the ground-up is setup to convince you that you are an individual and you need to go find a job and work like everyone else (hopefully with hefty student loan and other debts) and in our industry be treated like cattle and be subjected to some crazy interview process where you have to hack on some sh*t no one cares about.
if you start thinking about yourself as a business your goals become more like “how do I make myself profitable, how do I get to the point where another business cannot effectively operate without my services.”
is this a lottery? maybe… this is easier said than done but I would rather spend my career in this chase than be an employee somewhere where 99% of my peers don’t know my name
That depends very much on your world view. If you get the job, it would imply you've just cost a lot of other people $100ks. That simply can't be true, because there's a small number of such jobs.
The only way this can possibly hold is if you hold that only the "best" candidate should get the job, and by "best" you mean: the one that gets most of the leetcode questions right. But then there's no us.
Edit: I'm not in favor of leetcode interviews, and I do understand that there's a bias in interviews (which won't go away when you drop the leetcode).
Leetcode feels meritocratic. I suspect many people (engineers perhaps more so) don't trust their ability to "size someone up" from 45 minutes of casual conversation alone. Or worse, they worry that they'll have to defend their hire/donthire based on what amounts to a casual conversation whereby they were trying to judge character, affability, confidence, drive, motivation....
> Meet someone who went to your Alma mater? Same gender? Same race? Give them the same question as everyone else, but hint them through it, ignore some syntax errors, and give them a strong hire for “communication” when they didn’t even implement the optimal approach…
I do not know how hiring is done at megacorps; but on the few occasions I participated in tech interviews, we were looking for candidates to join the team that I was on. So, what I was interested about the candidates was how well they were prepared for the kind of work our team does; and how much I would like to work with them. I am sure this second question introduces all sorts of subjective biases; but then, hey, aren't you going to spend countless future hours with that person? At least the first concern incentivised us to make sure candidates were sufficiently competent. BTW, we didn't ask leetcode-type questions.
> but to the individual, this is costing us 100s of thousands of dollars, because there’s only a handful of companies that pay well and they all have the same interview process
I am not sure I am following the argument. If there is only a handful of companies that pay well, and if you only consider companies out of that list for employment, isn't it reasonable to predict that there will be lots of other candidates who want the same, far more than the number of available positions at those companies. How, then, should those companies select from such a pool of candidates?
> A way to suppress wages and job mobility for SWE. Who wants to switch jobs when it means studying for a month or two?
If what Google search tells me is true the average SWE tenure is only a couple of years or so.
I might need a month or two of study (or more!) to be able to handle the harder Leetcode problems, but if I was expecting to have to do this every couple of years or so I'd consider taking some steps to maintain that knowledge between jobs.
That would probably be considerably less overall effort than forgetting it and having to cram for months every couple of years.
And you think without leetcode questions the despicable biased interviewer from your example would not be able to hire someone from their Alma mater over someone else?
Casual gamedev here.
Ive written like 100 quadtrees from scratch so its a goto algorithm Ive got down by memory. I used to think this datastructure was very basic and common because young gamedevs often share it online with eachother. Highschoolers etc.
One time in a technical interview I realized a quadtree would solve the problem they were asking. I also happened to be reading SICP and practical common lisp on the plane on the way over to the interview. (specifically the sql sdl macro chapter) Unfortunately this gave me an idea. I narrated, sketched out and pseudocoded a plist based quadtree on the whiteboard and felt pretty good about myself.
I turned around to see my interviewers were not impressed. Actually they thought I was an idiot. They didnt know what a quadtree was, and they were quite upset I used "tuples".
I assured them its very efficient, and that the compiler would take care of "the tuples".
They failed me. Moral of the story don't use common lisp in a javascript interview.
> A way to suppress wages
> A way to mask bias
> Same gender? Same race?
> someone you don’t like
Ok fair points. it really could be a thing when you've already decided and are just playing around the process. But ya'know even when there isn't an inherent bias from race, gender, alma maters and overall like ability, the whole tech interview process still sucks.
The shit soup presented at a coding interview, leetcode or take home style ones are a result of wholesale cargo cult stupidity than malice.
There's no incentive at any level in BigCo. Margins are too big for top level to care. its hard to measure, middle management layer can easily fudge metrics while driving the org effectively into the ground. Lower layer that made it through the filter, sick of this grind and are busy filing tasks sipping free coffee.
The elephant in the room is that you cannot make hiring a process driven activity.
This is org level dysfunction. There's more such exhibits - DEI, promo criteria, ERGs, re-org politics, vendor selection to name a few.
There's a difference between asking someone to invert a binary tree, and asking someone to sum up a list of numbers. The latter finds the people who can't code much more quickly!
This is so true. You see it when you are on the hiring end. There are all kinds of people out there. I fully expect every hireable programmer to be able to set up a class, instantiate it, and use types. A lot of people are not even at that level. They know just enough to be dangerous.
It depends how it is done. I used to think the same way, and I would never hire someone without having seen them program live.
But having experienced leetcode-style interviews on the candidate side, it's clear to me that they are no longer about figuring out and coding a solution on the spot. Interviewers expected a solution FAST, and to match that you need to have studied and learned the answer beforehand.
...Instead you end up hiring people who can memorize every test from leetcode.com without having the slightest idea on how or when to use those data structures and algorithms in the real world.
Nonsense. Ask a candidate to add a feature to an open source product similar to your own, and share the PR.
Ask them to do it with a time box.
Guaranteed better results.
No. The unfortunate reality is that if you don't do something - if you don't find a way to test whether they can actually program - then you will hire people who can't program. You need something.
But that "something" doesn't have to be leetcode. There are much better (and less abusive) ways of doing the same thing. All it takes is a programming problem that they've not seen before. That's it.
it sucks, but honestly having a very clear metric for success is significantly better than the wishy-washy interviews that predated this. i'd much rather grind and be prepared to just regurgitate some algo/code and explain it than deal with the color of the sky being wrong, interviewer woke up on wrong side of bed, i was coerced into doing a "project" that equated to just unpaid work, etc.
whether or not you succeed as an engineer generally falls down to whether or not you know how to code (pretty much anyone is good enough at this for most jobs), can you communicate effectively (covered during the behavioral portions of interviews), and do you work hard.
honestly i think it sucks, but in comparison we're pretty lucky. i think a lot of people in this world would be happy to grind a few hours a day for a couple of months to get paid 250k+ base
Leetcode interviews are in general just incredibly poor signal-to-ratio; especially with the advent of LLMs making it easier than ever to just cheat on the problem. I would just generally assume that any company with that as an initial barrier isn't actually looking to hire anyone at all, but for the sake of 'appearing' to have employee growth.
I’ve seen many companies try to do something more relevant in the technical interview, it still suffers from “which data structure manipulation method happens to be available in this framework and have you seen it recently”
Regarding “cheating” with an LLM, there is not unanimous consensus on that either. I’ve been in technical interviews where the collaborative coding tool had an LLM built right in, and I was encouraged to use it. If thats what I’m going to do at the company then demonstrating problem solving that way is useful.
I lead tech interviews for E5/6 level engs where I work.
Everyone knows how pointless it is. But it’s become more like paying the tax.
For the record, I typically give away the a-ha moment early on and see how the candidate is able to take the suggestion and work through a problem with me.
> I don’t really have a solution to this problem, I just know it’s a problem.
The solution is simple: refuse to do those interviews. I have, and I wish more engineers would. It's one of my screening questions: do you do whiteboard style quiz/brainteaser interviews? If the answer is affirmative, I politely (and sometimes impolitely) decline.
I've written a book, edited another, contributed to plenty of open source, including Golang, started a few companies, and been around the block. I'm more than happy with a take-home or with going over code I've previously written, but doing leetcode in a "live coding session"? I'm good, thanks.
For FAANG and other huge companies, it's probably a decent heuristic (though I doubt even that), but smaller startups or non-tech companies doing this nonsense is a clear red flag. I mean, there's entire books dedicated to gaming the software engineering interview. The signal-to-noise ratio is probably so terrible, it's essentially just professional hazing at this point.
I am starting a company, and I need smart people; I do not care how good they are at programming language X, or technology T - all the skills my employees will need can be learned on the job.
I want to optimize the time it takes for an arbitrary hire to become independent, to have learned the basics, and to make meaningful contributions. They would write code at most 20% of the time, and the job has many other nuances.
In your opinion, what would a process that you wouldn't refuse look like? Would a learning interview (I present something that you have not seen before, act as an oracle, provide docs, evaluate how quickly you pick up concepts) be so insulting?
> all the skills my employees will need can be learned on the job.
> optimize the time it takes for an arbitrary hire to become independent
So you are looking for somebody in top 5% cognitive ability and top 5% independence which is what pretty much every other employer wants.
My advice is if you can pay competitive FAANG salaries (or compensate lower salaries by prestige/saving humanity/bringing people to Mars), just do what everyone else is doing - Leetcode, behavior interview and architecture interview. Innovate on something else, standard interview process will give you smart motivated folks.
If you can’t pay those salaries, you need to be creative and decide what compromises you are ready to make and tailor interview process to them. No one size fits all process here. You are essentially looking for people who failed in standard process while still posses qualities you need. Expect to spend more time on hiring and have more bad hires.
realistically we're looking for much higher (sub 1%) cognitive ability. I understand this is highly sought after, so we incentivize by paying a few times what FAANG does locally.
I was genuinely wondering how OP preferred to be approached vis-a-vis this sort of assessment, since they suggested that they would walk out on conventional approaches. I mentioned cognitive ability and ability to learn as I feel those are harder to extrapolate from one's existing publications/contributions/take-home assessments, compared to in-person discussion(s).
In this particular discussion, I am not looking for the best process for the task. I am looking for a process that will not get deemed a "red flag" by candidates.
I have frankly dealt with some people, who wrote books, had lots of titles, yet were incompenent in the very subject their books are about. So yeah, does not mean much to me. I personally love whiteboard tests; shows your ability to code quickly and reliably.
> yet were incompenent in the very subject their books are about
Gonna <press X to doubt> on this one. I mean I guess anecdotes are cool, but yours goes against basically any editor/publisher due diligence, not to mention the financial incentive of writing a book in the first place (the book actually being good), but alright.
It does not have to be bestselling books; just some niche books. And also I can guarantee you I can write you a book, say, "intro in datascience" without actually knowing much about the subject, just by mixing together existing material; after that succesfully forget everything I had learned why was writing the book.
I see these posts a lot, and I sort of disagree with where they are coming from so I'll play devils advocate. Leetcode is used to select for one of (or both of these) 2 things for the average IC hire:
1. Do you work hard (i.e a person who studies many problems, memorizes solutions, and regurgitates then)
2. Are you smart (i.e a person who doesn't do any leetcode but knows the fundamentals and can sythesize past knowledge well enough to answer a new question)
Now in a vaccuum, my takeaway from someone who doesn't pass a leetcode question is that they are more likely to not be either of those (given no other information about them). In my opinion, solving at least one leetcode easy - medium question (maybe with some direction from the interviewer) should be the standard for any role where you need to code.
And realistically, companies index less on leetcode skill as you become more senior or apply to more niche roles. And plenty of companies out there have bug squash rounds, system design, behaviorals, take homes, trivia, background specific questions, etc. Leetcode is usually not the be all end all for hires above entry level. Leetcode is not meant to magically select for the best engineers. Its simply another signal for the hiring committee to use.
2. Are you smart (i.e a person who doesn't do any leetcode but knows the fundamentals and can sythesize past knowledge well enough to answer a new question)
Leetcode questions can by definition never come close to real life problems - they are simply out of context. Working on a project even a few weeks will grant a smart person any insights he might actually need. The kind of out of the blue puzzle problem a leetcode presents is more akin to crosswords in old leisure magazines.
> 1. Do you work hard (i.e a person who studies many problems, memorizes solutions, and regurgitates then)
This isn't working hard. And you are competing with the people that are working smart. Such as people having a separate computer/screen to either look up the answer to a problem or ask an LLM to regurgitate some code you can transcribe. It selects for people that are dishonest.
> 2. Are you smart (i.e a person who doesn't do any leetcode but knows the fundamentals and can sythesize past knowledge well enough to answer a new question)
See above. A lot of the problems have nothing to do with real world coding scenarios. Do I need to pull out memoization and know how to work with grammar rules off the top of my head? Or the functionality of quad trees and how it applies to partitioning? No; I've literally never used them in my career and likely never will. What I do need as a software engineer is understanding the high level concepts and when to apply them and refresh myself on the use case when and if appropriate.
It's skills that a 4-year college student learns at approximately 19 and then promptly does not use for their next 50 years of actually building things.
It leaves out looking at the engineering skills it takes to put something past the finish line on budget.
It's at best, an adjacent sideshow and at worse a proxy for ageism to preference people closer to 19 and less capable at getting quality software out the door.
And maybe that's why the industry has such a damn hard time getting quality software out the door.
Maybe.
Here's a far better idea. Give them a broken build or crashed code or some other failure and ask them to fix it. If they can navigate around a system, diagnose a defect, find the cause, offer a patch, and match the coding/commenting style etc, that's an actual player you need, not someone who can still solve an algo question from their CS101B midterm exam at 37.
This version can go deep as well. You can have separate branches and an issue tracker, maybe with the conversation of the bug already going. Maybe the comments have a patchfile that needs to be manually applied. Maybe the patch has a new bug in it ... Perhaps it's a red herring and addressing a similar but unrelated issue. Maybe it's a closed bug and a regression and nobody wrote a unit test for it. Maybe the unit test is there and it's up to the candidate to discover it or the unit test is out of date, buggy and useless.
This is the real stuff.
Get them to communicate after. Saying something like "the unit tests were broken" versus "I was unable to understand how to get them to work right so I moved on" ... That's a huge difference in attitude - only one of them is a team player.
also if you're going to "timebox" be very very generous. If you think it takes 30 minutes, give them 5 hours and say you're being generous.
I've known people who program things like jet engines and stuff that goes into space. They know their stuff, are super methodical and take their damn time to do it well.
Quality programming is a rarely practiced time consuming art and you'd be lucky if you have 10% of your workforce doing it. Heck, you'd be lucky if you have 1%. Those people save the ship.
I'm not one of them. When you meet them, they're like a different kind of human.
If you timebox them to 30 minutes, the people who write mission-critical quality code will fail to get it in on time and be filtered out.
These are the people who write those 10,000 word articles that appear on HN with carefully constructed schematics, equations, and code that you struggle to understand and bookmark to study later.
On the other hand, if you give people 5 hours, I assure you, the morons will still obviously be morons and you give the "True" Engineers who take their craft dead seriously a chance to shine.
Yeah, no. Show me your GitHub commit history, and point out where you used some leetcode magic. Should be easy, right? Like every other week you'd be using it? Or... is 95% of code somewhat boring, and ideally maintainable?
> 2. Are you smart (i.e a person who doesn't do any leetcode but knows the fundamentals and can sythesize past knowledge well enough to answer a new question)
I'm sorry but it's incredible hubris to think you'll come up with say Dijkstra's algorithm during the interview from the first principles. The only reason you can do it is you studied it before.
So someone coming up with the right answer and claiming they've seen it is useless.
Right answer and claiming they hadn't seen it is basically a guaranteed liar.
You can use it to filter out the liars I guess. But you'll also filter out your Claude Shannons as well.
Either way, it's a dumb test unless you're genuinely getting a steady stream of Princeton and Cambridge magna cum laudes waltzing through the door and can reliable bump the last two priors by a few decimal points.
> 1. Do you work hard (i.e a person who studies many problems, memorizes solutions, and regurgitates then)
Most people applying already have a job. And if they work hard its actually quite uncommon for them to have ample energy to do leetcode questions on the side. So you might get the opposite - you'll get candidates who aren't great performers in their current roles but used their extra energy to grind leetcode.
I know people in the four quadrants of {no-hire, hire} x {would not pass LC, would pass LC}. In my world, conditioning on [would pass LC] does not improve the hire odds.
Even worse: it misses out on the hire people with no formal training too busy to waste time on skills with little use outside of the interview (in the real world, they'd use a library).
Spend enough time around hiring teams or recruiters and you'll see that 99% of them are simply very bad at hiring. They put almost no effort into a search. They can definitely find someone. But finding someone that is hired for their actual ability to solve the problems that the shop or team is working on is rare. Leetcode is just a way of culling the herd. The idea that it somehow reflects the requirements of the job are a useful fiction to obscure how little effort most teams and shops actually put into hiring.
>And realistically, companies index less on leetcode skill as you become more senior or apply to more niche roles.
when does this start? I'm over 8 years in and I'm tired of being told to make a dang NxM Bingo solver in 30 minutes.
>Leetcode is not meant to magically select for the best engineers.
I'd agree with you in 2022. in 2024, I'm being way more filtered out for questions that I get right but didn't solve fast enough, or with zero hints. I was never a leetcode savant (too many other things to study for, and LC is maybe 20% of my interviews) but I don't feel that much worse than 2 years ago. Companies just have sky high expectations and are on a bargain bin for 15+ year roles paying 5 year salaries.
And I am so sick of „I am so sick of leetcode-style interviews“-style posts. Yet here we are. How much more can this dead horse be beaten? Aren’t you tired of discussing this topic with the same arguments each time?
it's clearly not a dead horse as long as the leetcode interviews are still ubiquitous in the industry, and as long as people agree it is a problem. the posts are trying to convert a critical mass of people into realising the problem exists, and that they might be able to be part of the solution.
The last few interviews I've run, I've shown the candidate sample code and asked them to review it. We have samples that have at least one bug / problem per line. Then some modeling task (whiteboard). We have some simple exercises (fizzbuzz style) to figure out if juniors actually can junior and as a springboard into discussing topics like runtime complexity, memory layout and stuff. This is all highly interactive, depending on how well the candidate does. The idea is to figure out if the candidate somehow faked themselves through screening as well as to somehow get a feeling for cultural fit...
I'd give a bit of guidance. This can go so many ways. I can see many right but totally different answers.
In any interview, I kind of presume that the interviewer has a "perfect answer" in mind and is playing some game where I basically need to guess what number they're thinking of.
I'd make it clear that anything is ok. Just go some direction
Oh definitely. I start with very basic syntax level changes. I want to hear the candidate say "switch case or if/else tree" within the next 20 mins at the least (for SDE I).
And the problem can be scaled to any complexity. Distributed/Design/Data Storage
Looks great and will give it a try in the future. Refactoring is one of the fun things in our job so making it as part of the interview is great.
When I make interviews I also come with the code, normally a few line snippet and have the person tell me what it does or whether there might be a bug and talk about it.
Reviews are also great. They take away the show-cooking aspect where many people feel nervous and makes it more conversational. It is also very easy to evaluate as well because everybody debugs code at some point.
I personally dislike even fizzbuzz because there are always clever ways to go around it that are totally irrelevant.
I was once asked to implement sort a list in python, so I just took the list and passed them to sort. But no the interviewer wanted me to implement a specific type of sort…this was for a distribution maintainer job and I almost failed it because of that. Fortunately the other stages were actually technology related and I aced them.
Send the candidate off with a broken unit test to fix, own tools, comfortable environment like home, 1 hour time box.
I like to do something horrible based on timezones, or phone number validation, with a Wikipedia article that is the spec.
At the end, ask the candidate about their solution, and do a code review session.
The (best) seniors just use a library.
The good candidates roll their own, but can explain the tradeoffs.
The juniors get the test to pass and stop.
> something horrible based on timezones, or phone number validation
What? Do you have something against zip codes and email addresses?
I like to ask something to do with strings - English sentences. I tell them it's intractable in the time we have and that I really want to see how they whitebox-test the code they write, so take ~10m timebox to write the code, then write some tests - at least one test that finds a bug.
likely not as there are other, more efficient ways to roughly gauge that.
It does however test one's conformity to arbitrary processes to some extent albeit in not the most efficient manner as well...
"Leetcode questions are the worst form of interview questions, except for all the others" - Churchill
I too used to hate leetcode questions, but after conducting 100s of interviews as an interviewer, leetcode style questions are the best way to test intelligence and coding skills ive found in general.
However, leetcoding interviews must be paired with system design and topgrading in order to have balanced interviews. Leetcoding alone is incomplete but i find it essential.
It's true that nobody is expected to balance a binary tree as part of the job but the point of these questions is to see how you approach the problem and how you communicate your solution. Given that you can't perfectly predict how someone will do at the job, employers use leetcode problems as a proxy. Even those who memorize leetcode solutions must also memorize how it works and understand the solution. Given that the problem is random and you're likely given multiple interviews, it's unlikely you've memorized the exact problem and solution without cheating. If you've memorized enough solutions that it's likely you've seen the problem before and you can understand the solution enough to present it, then you're in the 0.1% and deserve to pass.
> the point of these questions is to see how you approach the problem and how you communicate your solution
I disagree. It’s supposed to be that, but the reality of the interview is that there isn’t any time to problem solve so you need to learn the problems and just regurgitate the solution.
That doesn’t put you in the top 0.1%, it ensures you’ve spent dozens of hours practicing questions.
I understand the frustration, but at the end of the day I think they do serve a purpose. They're not great at that purpose, but they're good enough and they generally produce false negatives (smart people fail) rather than false positives (you hire an idiot). And this makes sense, because a false negative is low cost (maybe you spend 50% longer interviewing candidates) but a false positive is high cost (you hire an idiot and have to spend months establishing that, firing them and starting the hiring process again).
The side effect of this though, is that it's an extremely painful process for the applicant. Even if you're a great software engineer, you're going to freeze up, miss something, have a bad day, and fail a few interviews. Now for most software engineers failure is an unusual and harrowing experience, so they react really badly to it. Especially ina
scenario where it's difficult to blame anyone but yourself. But just don't! It's fine! just move on, there's plenty of jobs and the interview process is a blunt tool, not a final evaluation of your worth in life.
> a bad hire can be an existential threat to the team or the entire company.
Totally, but leetcode wont filter them out.
Incompetent is fairly easy to navigate around, pathological bastardness is impossible to work past. I know that some companies use psychometric tests, but they are not based in anything remotely scientific, but are also super easy to game.
> but coding challenges do catch a lot of bad hires.
leetcode doesn't catch bad hires, as someone whos done a lot of interviews, we could easily catch "bad" coders with a very simple whiteboard test or a "look at this code and tell me what you think"
but people being bad at code aren't the company killers, its the toxic people who are "good" at coding that do. They can be very good at these kind of tests, but are absolute shites to work with.
I think we actually agree more that I let on, and you are right that coding tests should only be a part of the interview process.
You are right, there's not enough time to filter a bad hire. Not enough time to fully vet anyone during an interview. We can work a challenge together to see how you work collaboratively
"At-will", he says to the expat to Finland. I wish. It's nearly impossible for me, or any FTE I know here, to get fired. (To be fair, I have never to my knowledge been considered 'a bad hire' by anyone except Home Depot.)
But in fact even in at-will situations, firing aversion seems to be a real and expensive issue that all kinds of managers in organizations face [1] [2].
So, in one sense I agree, bad hires are existential threats "only" because we don't fire them fast enough. On the other hand, have you ever had to fire someone yourself? Try it. It's heart-wrenching, even to a heartless utility monster like yours truly.
How do you know they produce false negatives over false positives? How do you know if your top candidates simply didn't cheat over the honest candidates that produced worse but honest code?
I’ve interviewed people who wore an earpiece and were fed instructions on the fly, and also people who kept their cameras turned off to try and hide the fact they were getting assistance. You started picking up on patterns like them repeating your questions, or spending a lot of time in silence (asking their friend a question) without any other communication or activity on the screen. Like they weren’t there.
I once interviewed someone who was on a call with someone but messed up their audio feed. I could hear the third person but the person I interviewed could not.
There's an entire industry of bootcamps built on the premise of teaching you how to pass this exact test in like 6 weeks.
And because they ask pretty irrelevant questions, the 6 weeker will do better then the recent Stanford PhD who has been studying a specific problem for 5 years or the multimillionaire who spent the past 10 years building and selling successful software companies.
Those people will fail and your 6 weeker who's never heard of cmake, diff or gdb will pass.
When I see there's leetcode as the hiring process I presume the place is full of bozos because the process actually strongly preferences them.
I believe this is a common myth. I've spend around seven years teaching programming and people who can go from zero to decent problem solving in just six weeks are virtually non-existent.
Either it wasn't a true "zero" but rather they were really good e.g. at advanced math or physics and managed to see some parallels, or it has to be a genius. I'd be very interested in hiring such a person.
Leet code is not "decent problem solving skills" but it's almost always about three types of programs: sliding window, linked list rearrangement, and BFS.
The code to solve basically all of them look nearly identical after you classify it and the questions asked (what's the big O of this) are also all identical.
It's an extremely narrow and teachable class of problems.
There's outliers for sure but the 80th percentile are basically just the same damn things slightly rephrased. It's extremely gameable and doesn't test the kind of breadth say, fixing a crashing program would show.
If you think someone passing that means they can fix the conda bug in your k8s cluster or whatever real problem you have, dream on. It'd be like hiring a cook by quizzing them on the names of kitchen utensils.
Well the thing is, I keep hearing about these people who can't code but excel at leetcode, even that they are mass-trained in some bootcamps, but somehow I never saw one in reality and I interviewed hundreds of people. So I'm kind skeptical this is a real issue.
The other way round (people who code, but can't leetcode) I agree it's an issue. Unfortunately as noted here in comments, cost of a bad hire is much much worse than cost of a no-hire, so it's the price I'm willing to pay.
I think all leetcode gives you is a proxy for IQ in the language of code. If someone excels at leetcode, they should be relatively smart and able to write code.
Excellent! That’s a something, a something you can measure. Back when I was involved in hiring, it was definitely better than not having that (at junior levels).
However, the idea that “IQ is all you need” is so obviously false that I think you need to reconsider. Bill Gates famously, (infamously?), hired for IQ. Microsoft used puzzles as their IQ proxy, and I think leetcode is just a modern variant. Possibly a worse version, as training has a bigger effect.
Bill Gates later would claim that he overvalued intelligence. I would say that intelligence is necessary but insufficient. We would get further, faster, with a quick IQ test and then dig into the candidates experience, communication ability and personality. No need to study!
The FAANG companies all have a line a mile long of candidates willing to put up with any amount of BS for a chance to work for the name and $$. If you aren’t such a company, trying to hire that game changing person for $$’s you can afford to pay is going to require a different approach.
> I keep hearing about these people who can't code but excel at leetcode
I think the mass trained at bootcamp is an exaggeration. The best bootcampers are those that use it as an extended education from learning the traditional ways. I'm sure most people here complaining about leetcode can indeed spend 6 weeks there and be interview ready. But few have that time nor spare cash lying around for such a venture. And why should we need to spend thousands just to ace a quiz like it's some SAT?
>cost of a bad hire is much much worse than cost of a no-hire, so it's the price I'm willing to pay.
At this point I wonder. Hearing way more stories about "survivors" from layoffs burning out from jobs that keep promising that they'll hire more staff to help out with. When in reality they have a hiring freeze or are trying to offshore the work that ends up needing more time to be fixed.
The stalls in a no-hire aren't just costing money in the lack of productivity, it's costing currently working employees who are doing 2,4, 8+ jobs without any meaningful increase in role/salary. It's not sustainable. You're draining a little bit of blood from a rock, but the rock is clearly starting to crumble.
No. The point is the correlation is poor between the two sets AND there's equally quantifiable things with higher correlation.
If I genuinely wanted a job again all I'd do is go to one of those drilling game websites for a few nights and then say whatever the interviewer is looking for.
Any test I can fail one day and pass 3 days later that is ostensibly supposed to assess whether I have skills that take 15 years to master is total bullshit.
What you complain about is a false-negative error: an engineer with 15 years of experience can't pass these tests unless he prepares for 3 more days. That can sometimes happen, yes.
The initial stated problem was different: false-positives, bad engineers practicing for 6 weeks will pass the test. That I seriously doubt is a real thing. If they managed to learn that fast, they must be good. They could be junior-good -- that's perfectly fine, you need to hire juniors too. You don't judge seniority solely based on leetcode results.
> Leet code is not "decent problem solving skills" but it's almost always about three types of programs: sliding window, linked list rearrangement, and BFS.
I could be wrong, but I think dynamic-programming problems belong in that list.
It's not "decent problem solving" that these folks demonstrate though, it's the ability to grasp and regurgitate common DSA patterns that show up in most leetcode style interviews. Does this ability make these people highly desirable hires? Should you hire such people over engineers with more real world knowledge, but who don't want to spend the time learning or practicing leetcode?
Well said, like IQ tests, if you train for it you're better at it, but it the meaning of the results is dubious.
Recently I failed the first round of a fintech analyst position because I didn't complete the 9 undergrad (or high-school, in certain talented schools) math problems (I have a math PhD) in the 1 hour time frame. In retrospect, I should've cheated using computer/online assistance, and my true failure was my unwillingness to cheat. Not because fintech analysts are cheaters, but because the testing process is nonsense, so recognizing that, one should cheat it (indeed, an analyst mindset!) I'd rather be jobless than to adopt that mentality though. (and in fact I am, the only job I was able to find was a minimum wage service-type job, it can't even pay for daycare, I'd rather raise my child on my own while wife works thank God.)
We have high-interest rates and global crises resulting in companies not hiring while simultaneously governments give orders to news media to tout the strong economy. Ineffective hiring practices will always exist, but people don't complain when jobs are available; you only see this type of discussion online when things have gone sideways.
Not my experience at all. Ask a bootcamper a slight variation on a problem they're clearly doing from memory and they fold. Ask a solid CS person, and they usually get intrigued, and dive right in.
They do produce plenty of false positives - just not "you hire an idiot" kind. You will hire a person completely unsuitable for the actual job, but is good at leet. Issue is not the lack of intelligence. Issue is not being software engineer despite being good at algorithms. Or not being good at whatever your position requires.
I do work in a team where majority was hired by puzzles of sorts. All of them are smart. They are not software engineers and it shows massively.
>They are not software engineers and it shows massively.
Coming from Real Engineering and making 2x more programming, I lol at people who say there is any Engineering in software.
We are programmers dealing with layers upon layers of abstraction. Knowing to optimize for time by using vectors is more of an art, than a science.
I did do safety critical C code and Assembly which are probably my hole in the whole 'its not engineering'. But javascript/python/backend work: programming.
The fact that in 99% of cases "software engineer" is just an aggrandizing title given by management to any and every programmer at their company should put a bullet in the head of the pretentions of this being an engineering field.
The term is used ubiquitously in the field of technology and has been for a long time. We also have Tech or Software Architects, and quite obviously they don't do Real Architecture™ or civil engineering, but the concept makes sense.
It's just a job title, arguably one without the protections or standards afforded to chartered engineers, and with an incredibly low barrier to entry. Pretty much the reason we have to do this ridiculous dance with tech tests, LeetCode, and 12 stage interview processes.
Hillel Wayne did a good "study" on this by speaking to a bunch of engineers who moved into software, engineers who stayed in software, and engineers who only worked in software (yes, you can be a chartered engineer in most Western countries including the US and UK purely from software).
The strong consensus was that software is an engineering discipline, and the nitpicking you can have on any particular topic that is/isn't engineering-y can be applied across the board.
But this conversation is totally moot IMO because if you have a bachelors in computer science or software engineering, and a masters in the same discipline, and several years experience, you can apply and become the same chartered engineer from the same association that does "real" engineering.
> The strong consensus was that software is an engineering discipline
I see it as a difference in time and stakes.
We’ve built bridges and boats for thousands of years, where the outcome of failure is people die.
This has a lot to do with why it’s easy to estimate the time and cost to build a house, but it’s a rare shaman who can consistently estimate software projects well.
Once we’ve built software for even a few hundred years, we’ll probably have it pretty dialed in, too.
> We’ve built bridges and boats for thousands of years, where the outcome of failure is people die.
We write software that has killed people due to bugs too, such as Therac-25 releasing too much beta radiation killing patients.
> This has a lot to do with why it’s easy to estimate the time and cost to build a house
My parents and in-laws are property developers and I've also been involved in a build; not one was on time or budget. Some are really off budget, beyond the scale software projects that mess up are. In China they've had projects run into the billions and be decades late.
I do understand your points, but this is what I mean when I say you can nitpick any issue with software not being Real Engineering and apply it to other engineering categories too.
But yeah we may have a very different process in 50 or 100 years. I'm only 10 years into my career and it's already pretty different to when I started!
One related problem I see is that Computer Science tends to be under-estimated by other engineering fields. A ton of software engineers did not study software. They learned to program as part of their engineering studies, and they believe that they learned both Computer Science and their engineering topic at the same time.
I may be biased, but I see fewer computer scientists pretend to be mechanical engineers because they once tuned a PID.
My mistake. If you wrote a random forest once, it certainly means that you should be the CTO of Google. And that chem engineers are all smarter than computer scientists.
>I lol at people who say there is any Engineering in software.
There definitely is. Safety/mission critical code is definately a more rigorous, formal process where you don't just download a library for padding a string (you probably wouldn't use built-in strings anyway on this kind of code).
But I agree that the way most coding works (i.e. move fast and fail often) is completely antithetical to how engineers should work. But there's infinite moneys and almost zero accountability, so who's gonna stop them?
>Knowing to optimize for time by using vectors is more of an art, than a science.
that's definitely why I fully agree with calling programmers creatives over engineers. once you get past dead simple problems, no two programmers are going to approach a prompt the exact same way. performance, scalabiliy, documenting, version control friendliness, even ephemeral stuff like code style. coders very much are artists who happen to know a lot of math.
As someone who started going to school for Mechanical Engineering and ended up in Comp Sci, I think your last paragraph is what drew me in. So many different ways to solve the problem at hand.
The fit/culture/soft skills/ is actually a fall back to "where did you go to college and tell me about how good you are", which was the pre-leetcode standard.
Real-world problem, likely related to what you do. Simulate day-to-day problems. E.g. you are a marketplace business, ask them to do a simple basket system, or a checkout or a product catalogue etc. Ask them to review some code and provide their analysis.
FWIW I think LeetCode problems can make sense for some businesses, but most not.
Hard leetcode questions hire grinders, not thinkers. Many of those questions were thesis/publication-worthy decades ago when first solved; it's unlikely for even a very intelligent person to produce a similar result in just tens of minutes. So if someone solves it, it's much more likely to be because they saw it or a similar question while grinding practice questions than that they genuinely derived the optimal solution from scratch.
I disagree with this. I can solve the majority of leetcode hard problems in under an hour despite not having seen them before and I'm far from the smartest person I've met.
Probably because you said you "can solve the majority of leetcode hard problems in under an hour". That could easily give the impression that you have practiced them. If you haven't practiced them, then how do you know you can solve them?
Under an hour is... not enough time for an interview. Needs to be under 20-30 minutes given that most interviews are 60 minutes and the other 20-50% is taken up by other conversation.
I don't know you or the people you've met, but if you can do this without studying the questions before and the solutions are performant, then I would put you in the "very smart" category
If someone claims they can solve Leetcode hard problems without studying the questions and the solutions beforehand, then I would put them in the "full of shit" category
You had to know this was going to get a lot of downvotes.
However, I have a real question as a follow-up: Have you been able to profit from this skill? If true, you should be able to crush any "whiteboard" programming interview. If I had it, I would move every 2-4 years and get large pay rises. You should definitely be able to pass FAANG interviews with that skill.
> Many of those questions were thesis/publication-worthy decades ago when first solved
The hare and tortoise (cycle detection via Floyd's) is a legendary impossible interview question. It wasn't solved until the 1960s with a PhD thesis. It is a hard problem to face in an interview. I had seen it twice in my career. I failed the first time. The second time I had the answer memorised, then put on my very best Morgan Freeman stage acting to pretend I was discovering the answer during the interview. What a joke. I was laughing to myself as I was "stumbling" towards the answer while the interviewer was providing minor hints.
If, after all of the normal interviews, you randomly selected candidates who passed and do not make them offers . . . you generally produce false negatives (assuming the interviews mostly worked).
> They're not great at that purpose, but they're good enough and they generally produce false negatives (smart people fail) rather than false positives (you hire an idiot).
I've seen this reasoning, but I'm wondering if this is actually true, especially if you also believe the other myth in hiring: most people applying are not qualified for the job. Since most hiring actually requires you to fill the position in the end (in other words, you can't leave it open indefinitely if no one qualifies as eventually the pressure to hire mount up), every false negatives does increase the odd of you hiring a false positives as well. Just take the example at the extreme, if there is only ever one person on Earth qualified for your job and you got a false negatives on him, there is no other way but getting an idiot in the job.
In my years of hiring experience, it doesn't hold up. I've hired folks who could leet code all day long, but couldn't develop a feature from scratch, or figure things out without enormous hand holding. I want to hire problem solvers who also know how to code. I stopped doing leet-code interviews about 7 years ago. By all means have people code in interviews, but don’t do leet code.
I'm retired now, but I used a variant of this that worked well for me. In my experience, much of what we do as engineers is learn new tools and application domains and apply what we learn to solve problems. So I wanted to test how good a candidate is at leaning and whether they appear to enjoy it (good and enjoy often correlate).
We would work together at the whiteboard, with me teaching them the basics of our application domain and our internal distributed computation framework. Then there were a series of 5 increasingly difficult iterations we'd work through. First, do a simple calculation. Now there's a need for a more difficult computation. And an even harder one. Then explain that it needs to run over billions of log lines, and it's I/O bound, so let's make it faster. Finally, a problem that requires them to probe me deeply about the nature of the data and only needs a hand-wavy answer (calculate a median in a single pass with limited memory - not too bad once you extract from me that the samples are all integers between 0 and 500). To the candidate, this looked like iteration driven by evolving requirements, not 5 interview questions.
I mostly let them drive, but I jump in now and then to keep things on pace. And I would try to answer any question the same way I would as a teammate / mentor.
They don't need to get every problem right. The increasing difficulty is designed so that almost everyone gets something right and something wrong (< 1% gets all the way through the last one). That's an opportunity for me to give supportive feedback to see how they respond. The best responses to feedback yielded something like "oh cool, I can also use that over here to simplify the code". The worst was someone insisting that you can flip the order of addition and division without changing the answer (i.e., order of operations doesn't matter), me saying "I think it does matter, try doing 1/2 + 1/4 in different orders", and them saying "no I have a math degree, I know what I'm talking about."
What they do need to do is be able to solve problems collaboratively, extract learning from an eager mentor, and extrapolate from those learnings to solve problems. The best candidates start finding trade-offs across the application domain / tech tooling boundary that I haven't even brought up.
One of the downsides is that it takes a while to calibrate as an interviewer, and that calibration is not entirely transferrable across interviewers, as candidates will respond differently to different interviewers. I did this interview thousands of times (my record was 14 times in one day, and frequently 20+ times a week - we were scaling an org from ~50 engineers to over 1,000). I never stopped learning about signals I was seeing, but I was mostly calibrated after a few dozen candidates. Leet code calibration is much more transferrable, but what use is that if it's not measuring what matters?
One thing that I think leet code interviews get very wrong is that interviews involve two sides trying to evaluate each other. Wearing people down with leet code is not a good pitch for your work environment. Many candidates told me at the end of my interview that they were shocked at how much fun/educational an interview could be, and many hires said that this interview style was the reason they accepted our offer over more lucrative offers, because they felt they would learn more and have more fun with us.
> One thing that I think leet code interviews get very wrong is that interviews involve two sides trying to evaluate each other. Wearing people down with leet code is not a good pitch for your work environment.
This is huge and is underrated. Even when I have (_rarely_) passed the most bruising interview process, my emotional state was so negative towards the company. Part of the OP's emotional state (exhausted / burned-out) is the result of an adversarial interviewing style. With your style, you can gently push people to the edge and see how they perform. Your style is more dynamic, which more closely matches real world working conditions.
Were those juniors with no experience developing features? Leetcode doesn't test if they have experience, you need to check both if you want someone that can do everything.
Someone who does well on leetcode is usually easy to teach so they will become good/great in a year or two, but if you don't have that time then go for leetcode+experience just like everyone else is.
Leetcode deselected too many good experienced candidates, which is really who I prefer to hire. I have hired good leetcode juniors that did not improve and are no longer working in software.
Did you select candidates with easy / medium / or hard LeetCode problems? Honestly, without studying, people need to be top 5% of candidates to solve hard problems. You can find real algorithms stars with those problems, but you will also pass on a large percentage of "good enough" candidates who can do a medium problem with a bit of hinting.
> if you also believe the other myth in hiring: most people applying are not qualified for the job.
Depends where you’re looking in the hiring pipeline. If you’re looking at raw applications submitted to a publicly listed job posting, I can say without reservation that the majority of candidates are not qualified. The pre-screening candidate pool is extremely bad on average, especially for remote job listings. Some people spend all day clicking the apply button on every job listing they see.
Your post constrains another misunderstanding about the hiring process: The goal isn’t to hire the first person through the system who is good enough to pass some arbitrary threshold. The goal is collect a number of candidates and hire the best one. People hate this reality, but it’s true.
> Your post constrains another misunderstanding about the hiring process: The goal isn’t to hire the first person through the system who is good enough to pass some arbitrary threshold. The goal is collect a number of candidates and hire the best one. People hate this reality, but it’s true.
I definitely do not have any misunderstanding here, and my description of hiring is the same as what you wrote, just in different phrasing: after a period of time, the company has to pick a candidate, presumably the best one by their interview signals.
I am disputing the standard claim of "it's better to have false negatives than false positives" by pointing out how that mindset does not work: since you have to choose someone anyway, the whole "avoiding false positives" might not be a thing at all
Okay, do do quizzes, but let people google goddamn stuff because that's just how software engineers operate. Don't make it an exam. Don't mimic the worst thing about the education systems. Both in the school and in the university, these goddamn exams were the worst because they tested memory first and everyone else second, and I'm such a kind of person that I could never remember things on command. It was a real struggle. I can't be the only one.
I've never been through the "regular" IT hiring process myself, but I've interviewed several candidates for an Android developer position. I liked asking questions and looking at how the person thinks. I didn't expect exact correct answers for every question, I just wanted to see whether they have a fundamental understanding of Android. After that, they had to build a small app at home (not my initiative) to show their skills.
I like this idea as the more advanced version of whiteboard programming from the year 2000. Even if you still do whiteboard programming, give them a PC / tablet, and allow them to do some Google searches. Then, you could even make the problem a bit harder to solve. Again: You are looking at the candidate's problem solving process holistically. I also very much agree with your second paragraph. Some of the most pleasant surprises I had during interviews were candidates who had very limited domain knowledge, but tried as hard to possible to make educated guesses, or ask intelligent questions. It shows a lot about a candidate.
What are you going to google though? For most of these LC questions you only need to know how to use arrays, maps and loops. If I'm interviewing you and you tell me you need to google how to append something to an array or check if the element is in a map then I'll think you just haven't written enough code yet.
There are better questions where I do think google should be allowed but they aren't LC-style. Some examples are building APIs or operating on files
At all the companies I’ve run we’ve used a simple whiteboard example for this reason. Don’t worry about typos; how does the person think.
“Trick” questions are dumb, but you want to get an idea of if the candidate knows their stuff (and let the candidate know what we’re like — interviewing is a sales process in both directions). If someone asks a question like “is it ok to modify the argument” (or says “I’ll assume I can’t modify the argument”) that’s great (or says “I can’t remember if strlen includes the 0 byte so I’ll add 1 — normally I’d look it up”) - again, no trick questions.
When I say “simple” it’s something like atoi or “shuffle a deck of cards”: basic, but not 100% trivial like fizzbuzz.
On of our best hires was a guy who made a fundamental mistake — when we asked him if it world work as intended he said “I’m a fucking idiot” and fixed it. You’re not supposed to talk like that in an interview I supposed but we read it as a strong positive signal.
We had one candidate who insisted that parsing a string and returning an integer was an unreasonable question because there was already a library function for it. “But what if you have an embedded system and no library?” (an actual situation for us, for part of our system). He was adamant — honestly that was a failure of the phone screen: had it been caught it would have saved us, and him, from wasting time.
I have come to like this approach reasonably well. My current company uses a similarly simple problem to evaluate candidates. For me, it was a breeze, and I thought we’d be hiring slouches left and right. Once I started administering the interview, I realized that the majority of candidates absolutely bomb this simple question for one reason or another.
It seems that “able to code a simple problem well” is a far less ridiculous proxy for good software engineer than “able to code a hard problem at all”. Much to my surprise.
“Simple problems” often aren’t as “simple” as the retrospectively sound if you consider everything like stress level, mixed mindset of the interviewee (they’re not just thinking about a programming problem, they’re also thinking about social queues, mixing conversing in, peer judgement, and loads of other anxiety inducing factors), etc.
The software industry is notorious for underestimating overall complexity factors, understanding uncertainty, and time to manage them and interviews seem to be no different. So “simple” problems are often good (I agree with you) because they’re actually not so simple for most people when you factor everything in, they’re often reasonably challenging under the set of circumstances.
Meanwhile genuinely difficult problems (in any arbitrary setting and longer timelines) require a whole lot of rote training and reusing little clever idiosyncratic techniques that may or may not be generally useful, so you very often end up with a question lottery on whether you already know how to address the problem in question or not correctly vs having some problem where a reasonable solution and expectations around that can be derived in the time (and environment) it’s taking place. Lots of nonsense.
Truth is, I'm never looking for "the best", I'm looking for "good enough to do the job competently". Because we can hem and haw over what is best for hours. And we could disagree with what qualities in what proportion make someone "the best". But we can all generally agree with "could do the job". As I've mentioned elsewhere, we're hiring right now and we have the duality of candidates in our pool.
There's one guy who is just a hard no from at least two people in the group. And that translates into a no from the third member simply because we're so adamant that he's not the guy. And there's another who we're all feeling really solid on. Not "the best", but solid floor and we feel they'll be able to learn.
But being able to weed out people who can't even perform the basics despite having a decent resume is invaluable.
I have a friend who has interviewed about 30-40 software engineering candidates for his company (small, publicly traded company from the midwest). He said that FizzBuzz alone is enough to filter out 25% of applicants. The next slight more challenging question will axe another 30-40% of candidates. I think Leetcode hards exist to stroke the egos of select software engineers.
Yea, smaller and medium sized companies just don't have the recruiting infrastructure to weed out the fakers. When I worked in FAANG companies, by the time a candidate got to me, they were pretty good. They went through enough filters and checkpoints and annoying gates that I never really had to do a FizzBuzz equivalent.
When I worked for more medium sized companies, I'd get software engineering candidates who obviously did not even remotely know how to code. Imagine the worst possible software candidate, and then lower your expectation even further. One couldn't even tell me what an "int" was. I think a single "for loop on a whiteboard" question would filter out at least 50% of candidates.
Well, I'm not impartial... I kind of have a chip on my shoulder against these phonies, but if you want me to speculate: they are faking their way through university the same way they fake their way through everything else in their lives: Through things like charm, good looks, Ivy League mannerisms, confidence, a firm handshake, a beautiful smile, that certain way of talking like a salesman. During Character Creation, they put all their skill points into "charisma" and just use that to sail through situations the rest of us have to work through. I think we all encounter these people all the time during our careers, and it's infuriating that what they are doing is reliable and works.
"You can lead a horse to water but you can't make it drink."
It says barely anything about the universities and everything about the student. If a student just wants to just have a perfunctory gargle at the fountain of knowledge instead of drinking as deeply as they can, that's on them.
Almost everyone embellishes their resume. Some do it to an extreme. The first stage should be weeding these people out, and any hiring manager learns this lesson very quickly.
Story: I learned this as a hiring manager early on. I was the resume screener, first contact and then last contact. Everyone in between was my team and other adjacent teams. I failed to ask the right questions in a phone screen and when we brought this guy on-site he failed very bad. Even worse, he broke down begging for for another problem he could solve.
It was bad for everyone and we were a very accommodating group that gives many tries. We weren't looking for perfect answers to one question.
That I wouldnt mind. I didnt do a full on CS degree, that never stopped me from being a good contributor to various teams over 8 years. I have learned the best candidates are the ones who like you say are open about how they messed up, and can get along with your team. It takes one toxic dev to ruin productivity for an entire team.
I agree with all of this. Leetcode questions are fine as long as they aren't actually difficult maths puzzles or brain teasers. We also used atoi/itoa loads and it was pretty much the perfect difficulty.
Though I still do start with a short fizzbuzz-level question because you would be surprised how many people fail that and it makes it way less awkward if you start with atoi and they can't write a loop.
If they fail fizzbuzz I generally just turn the interview into a chat or give them easy questions to practice, rather than say "yeah no thanks", because I feel like the latter is a bit mean, and everyone can benefit from interview practice.
> If they fail fizzbuzz I generally just turn the interview into a chat or give them easy questions to practice, rather than say "yeah no thanks", because I feel like the latter is a bit mean, and everyone can benefit from interview practice.
It’s kind to politely ramp the interview down in that case but I’ve learned to then cut off any later interviews. It’s a waste of the company’s time and the candidate’s too.
If we were going to have lunch the recruiter takes them to lunch anyway, unless the (now former) candidate doesn’t want to.
Do you mean the C function atoi() to convert a string to number? If yes, I also used the same technical interview question for year. It is a nice "fractally complex" (many levels of depth) question. Everyone can understand it in 10 seconds, but you can really show your experience when you get into the edge cases and unit tests. Plus, if the candidate is a bit junior, you can push their boundaries by asking about different edge case and watch how they react. Similarly: I had a teammate for a few years that asked people to program a very simple linked list class. He was shocked how many mid-level devs couldn't do it. As I recall, he never sent me a bad candidate, so it was good enough for us.
> We had one candidate who insisted that parsing a string and returning an integer was an unreasonable question because there was already a library function for it. “But what if you have an embedded system and no library?” (an actual situation for us, for part of our system). He was adamant — honestly that was a failure of the phone screen: had it been caught it would have saved us, and him, from wasting time.
Trying to debate the interview format or reject interview questions is one of my hard-stop rejection triggers in interviews now.
The couple times I’ve been part of hiring pipelines where candidates argued about the interview itself and then got hired anyway, they became extremely difficult employees within the company. Arguing about interview questions turned into arguing about every other ticket, code reviews, and every architecture choice.
I’ve had the same experience where we take an actual problem we solved in our codebase and turned it into a small interview problem. Same thing: Some candidates will try to debate the relevance or try to wriggle out of it.
there is a lot of stupid bullshit in most orgs, and things only get fixed by often irritable, unreasonable people. "shut up and do it" was only an acceptable answer when I was in the military.
if the interviewee pushes back and find they don't like the answer, then it's a good signal to both parties.
This is definitely something that is dependent on the company culture and the personal preferences of the team. I would never want to work with someone who is constantly argumentative over everything, just because he thinks he's pushing back on bullshit.
Sometimes you just have to "document your objection, but do it anyway."
Offtopic, but there's an interesting parallel between making a mistake like that and AI hallucination. In a human, that behavior is a positive signal, but for many people, that same behavior is proof of how LLMs are just a toy that will never rival human intelligence.
The difference is that an LLM isn't very good at saying "I'm a fucking idiot" and changing it when asked to double-check (unless you handhold it in the direction of the exact error it's meant to be looking for). Humans recognize their own hallucinations. There's not really any promising work towards getting AI to do the same.
Fwiw this is surprisingly subtle and non-trivial to get right. A naive "random swap" approach does not produce uniform permutations. Sorting based on a random comparison function can likewise produce biased results depending on the implementation of the sort. Tagging each element with a random number and then sorting based on the tag works (or so I think?) but you need to deal with the case where tags collide. Fisher-yates shuffle is bulletproof but I would not trust myself to write it from memory lest you end up writing Sattolo's algorithm instead.
It’s an open ended problem and someone could nerd out on the maths while another could define a card class, a suit class, and use a factory pattern (oh brother…SMH but probably wouldn’t categorically rule someone out for it, just talk some more).
At the end of the day the question is: “could this person do the job and could they work with others in the process?”
The problem isn’t about getting the perfect answer, it’s about observing the process, and conversing about the answer afterwards. I’d be fine hiring the candidate who whiffs the shuffle but knows about Fisher-Yeats and says they’d look for a well-tested implementation and drop it in. I mean that can’t be their answer to everything, but evidence of real experience should grant a Mulligan now and then. Coding interviews cannot be code themselves.
>because a false negative is low cost (maybe you spend 50% longer interviewing candidates)
this would make sense if they filled these positions in 2,3 weeks tops. But that's the insane part; I hear of interviews going on for 6-7 stages now, lasting almost 6-10 weeks. At that point I feel you are no longer avoiding false positives, but simply hiring the most desperate programmers willing to jump the hoops. The best candidates will get an offer mid interview at that rate.
>But just don't! It's fine! just move on,
I've been "just moving on" for some 9 months now. There's only so much morale in a tank, no matter how BS you know the whole process is. I just wanna move on in life, not be bogged down on how many permutations you can color an N x M grid with.
First, I am sympathetic to your experience. It must be difficult.
> I hear of interviews going on for 6-7 stages now, lasting almost 6-10 weeks.
From this story, it is clear that your market is clearly in favour of the employer. One year ago, it was a different story: half the time. Another thing that will hurt: If you are amazing, that 6-10 weeks will suddenly become 1-2 weeks. In my experience, nothing hurries the interview process more than telling an employer about late round interviews (or offers) at another competitor.
That false negative can be pretty damn expensive if the good engineer you didn't hire could have saved your ass from the situation the less-good engineer you did hire couldn't handle.
The problem with Leetcode-style test nowadays is that there's too much cramming. Otherwise, leetcode-style interview has certain correlation with one's geekiness and raw talent. Just this week my team solved a long-tail performance issue by first building a simulation of a queuing system, and also built a customized Coffman–Graham algorithm to resolve the order of execution of a complex task graph. I would certainly appreciate that my team members just get it when someone mentions basic CS concepts like topological order in a graph or queuing theory.
FWIW, I have come across examples in the wild of engineers who were extraordinarily polished and facile at everything leetcode, and poor at computer science. Not many, but some. This is a somewhat recent phenomenon.
Best tests are simple and common computer science problems in algorithms and data structures with no tractable solutions. There is no facile answer, you truly have to reason from first principles for any particular scenario because there is no alternative.
The market is so bad, and most interview sucks one way or another. The main reason is interviewer have tens or hundreds application for same position.
I have just went through one interview that have 4 rounds.
1. half hour talk
2. 2 hours coding for several crud api.
3. system design project ( I spend 2 days on it. they told I have about a week on it)
4. another half hours with CTO.
and then got rejected. (The interviewers does not bother to send a reject letter. I asked them they told me the position is closed).
How generous. One job I asked for an update a few days after my assumedly last interview, 1 week afterwards, and then 3 week afterwards (right before the holidays)... even by the last response the recruiter simply said "we are still evaluating the best project to put you on" or something to that effect.
That was 8 months ago now. Going through 5 rounds of interviews into a rejection sucks, but at least be straight with me.
624 comments
[ 0.17 ms ] story [ 435 ms ] threadIt is really unlikely, given a leetcode hard level problem to solve within 25-35 minutes and you can google your way out of it unless you are googling for the solution.
Idk anymore either and I refuse to conduct these interviews. I generally prefer somewhat realistic problems that are multi-disciplenary but have ties to real world applications related to the role.
1) A way to suppress wages and job mobility for SWE. Who wants to switch jobs when it means studying for a month or two? Also, if you get unlucky and some try hard drops an atomic LC hard bomb on you now you have an entire company you can no longer apply to for a year.
2) A way to mask bias in the process while claiming that it’s a fair process because everyone has a clear/similar objective.
Meet someone who went to your Alma mater? Same gender? Same race? Give them the same question as everyone else, but hint them through it, ignore some syntax errors, and give them a strong hire for “communication” when they didn’t even implement the optimal approach…
Or is it someone you don’t like for X reason? Drop a leetcode hard on them and send them packing and just remain silent the entire interview.
To the company this is acceptable noise, but to the individual, this is costing us 100s of thousands of dollars, because there’s only a handful of companies that pay well and they all have the same interview process. Failing 3 interviews probably means you’re now out $200-300k of additional compensation from the top paying companies.
I’ve interviewed for and at FAANGs. I can’t believe the low bar of people that we’ve hired, while simultaneously seeing insane ridiculous quad tree/number theory type questions that have caused other great engineers to miss out on good opportunities.
Someone will reply to me “if you know how to problem solve you will always pass.” Ok, come interview with me and I will ask you verbatim one of those quad tree/number theory/inclusion exclusion principle questions and I’d love to see you squirm, meanwhile another candidate is asked a basic hash map question.
Is the culture just so broken at these companies that it’s hopeless to expect people NOT to blatantly exploit the system for their friends? Why don’t people get fired for doing that?
How is this any less discriminatory than any other assessment based interview where you need to prepare? Non-assessment based interviews end up being vibes based which is much more discriminatory.
For example, at my current job, we hire web developers with Rails experience. Our technical interview process consists of either a pair programming session or an async/take-home task (candidate’s choice) which requires the candidate to implement a small feature in a Rails codebase. We do have some room to improve on objective evaluation of the candidate’s performance, but there is a test suite and a rubric which we use to evaluate their work. None of this should require that the candidate study, unless they’re coming in to the interview without Rails experience.
If you happen to have Rails experience but it was several jobs ago, the task we give you is basic enough that you should be able to Google what you need during the task to refresh your memory fairly quickly. In fact, I did this when I applied, having not worked with Rails in a number of years.
Edit: My main point is, even if you technically do have to "study" (really, just Google a few basic Rails concepts) if you're rusty, everything you do is preparing you for the actual job. Studying how to implement a hashmap, or computing the longest palindrome from a string of characters, or whatever other harebrained problem FAANG etc want to ask, is 99% of the time not really helping you prepare for those jobs.
For any given job, I believe there’s a way to design an assessment that allows the candidate to demonstrate skills directly relevant to the job, even if the job doesn’t have requirements that are as specific as “Rails developer with a few years of experience.” I have yet to see a single job listing in several decades in the industry where I think to myself “ah yes, the only way to find qualified candidates for the job would be to ask them to implement a quadtree.”
And my "implement a quadtree" example was hyperbole; again I'm sorry if you didn't recognize it as such. I do suspect we have wildly different definitons of what questions qualify as "relevant to the position," though, so I'm not sure how far we're going to get in this discussion.
because most other assessment based interviews are based on what you do on the job, likely stuff you've done at previous jobs as well. Less prep time when you spent thousands of hours already as a career.
Yes. Because interviewing is
1. hard, but no company has proper full time proctors. So "expert interviewers" are a rarity
2. not standardized in the slightest. So your performance varies entirely by the interviewer, their style, and their mood that day.
3. some weird blind test where you hope you studied the right topics. You're not getting the best out of a candidate, you're basically hoping they read your mind and know exactly what you want them to say.
>Is the culture just so broken at these companies that it’s hopeless to expect people NOT to blatantly exploit the system for their friends? Why don’t people get fired for doing that?
Sure. but that's a universal problem. "It's who you know, not what you know" is advice that has spanned centuries. I can't even blame the modern tech industry for that one.
That + the above issues with interviewing mean you're always going to go with a referral over a random applicant.
That isn't always true. When I interview HR gives me the exact questions I'm allowed to ask. These are vetted both to prevent me from asking something illegal, and also by research to get the type of things useful for interviews. Sometimes it is annoying - you can easially finish the interview with a great score but I have no clue if you can write code or not. However we are carefully trained on how to ask the questions and how to grade them.
I was mostly referring to technical tests, but I understand there are definitely some set of questions you need to ask no matter what. I don't really knock recruiters too much for repeating the usual "are you authorized to work in the US" kinda stuff even though it is the first question on their job application.
2. All tech companies I’ve worked at since around 2015 have completely standardized interview questions, sometimes also hosted in a GitHub repo where any employee is free to comment or even open a PR to request a change. Every candidate gets the same questions. This thing where “faang” interviewers just pull a random LC question out of their ass is complete insanity to me. An organized set of questions takes a senior engineer like a week to organize and commit to… And if your questions can be memorized and recited by rote memory and the candidate can do well on it without the proctor knowing, then your questions are BAD or your proctor is a moron.
3. I’ve never worked anywhere where I would describe the interview like this. Even the startup where I was the first engineer and there was no formal “test”, it was an hour long chat with the technical cofounder where he grilled me about coding skills and past experience/accomplishments. I won’t take a job if the interview isn’t asking me things that focus on my existing experience and skills, it’s a red flag about the company culture in general.
As for your “universal problem,” I disagree with fatalistic takes where you just throw your hands up and say “whatever shall we do” all the while YOU are the one benefitting from the system that cannot be changed. This is how simple-minded people think about the world.
2&3. I wish I could relate. I work in games, fwiw, so maybe it's simply more wild west than the other domains in such regardss. It's not even that I can't answer some of the questions, but if I'm asked a question on CPU architecture after I spent my time studying a bunch of C/++ trivia, I'm not going to answer the former as confidently since I'm pulling out knoweldge from a 3 year memory bank that I happened to work with at work.
>all the while YOU are the one benefitting from the system that cannot be changed.
I am? I'm a year out of a job and I sure wish I had a referral. My last studio shut down so those references scattered all over.
But for "whatever shall we do"... I don't have some magic perfect interview, but a few basic steps to save everyone's time:
1. we don't need 5-10+ rounds of interviews and 3 months just to make sure a candidate isn't a lemon. IMO if it's over 3 rounds (not including an HR call for basic alignment), there's some process that really isn't necessary. You dont need to speak to a CEO unless you're applying as a director or executive yourself. You don't need to meet every team member separately for their variation of an interview
2. Ghost jobs absolutely need to be regulated. I'm surprised it's even legal to lie to your shareholders and say "we're always growing" when half your postings have little intention of attracting a candidate. At the very least a job posting should delineate between urgent (hired withing 1-2 weeks of response), "normal" (2-4 week process), and "cold" (6+ week process, non-necessary role).
3. jobs should in fact give a general direction of what kinds of technical questions you expect to be asked for that portion. Who does it really benefit if one candidate happened to study leetcode (and the right kind of leetcode), but the other was focusing on system design, or questions about their specific industry problems solved? Or domain specific probblems?
Just a few ways to be honest, efficient, and bring out the best in candidates. Interviews shouldn't be so hostile to the labor the company clearly requires.
> if your questions can be memorized and recited by rote memory and the candidate can do well on it without the proctor knowing, then your questions are BAD or your proctor is a moron.
I agree in spirit. My only caveat sympathizing with interviewers is that most questions can indeed be "memorized", and arguably all questions/practices can be studied for. And there is some merit to trying to figure out how someone approaches a new problem. I think that approach is highly inefficient when you do in fact mostly want someone who "already memorized" the right domain knowledge, but I digress.
But people and their experiences are diverse. Without knowing their background (and again, some people seem to read the resume for the first time during the interview) you are never going to surprise 100% of your candidates with the same question bank. But that may be too much of a burden on trying to cater every inerview for every promising candidaet.
How do you catch them? Unless they’re idiots they’ll write facially plausible explanations of their hire/no-hire decisions.
The problem isn't standardized tests, it's that leetcode questions are about having the time to have learnt the answer beforehand, rather than raw ability for problem solving.
LC tests typically copy problems from the university the interviewer graduated from. College programs differ, so this is really a case of what you were introduced to.
There's a fairly popular online LC test company in my corner of the world which was formed by graduates and lecturers from a certain university and they started out by just giving the problems from the curriculum. Result was heavy bias in favour of students and alumni of that university.
If that were the case then a normal, well accomplished software engineer shouldn't need to "grind" leetcode to pass an interview.
its just a cargo cult. Getting someone to do a code review is a much much better test of skill:
Do they ask questions?
are they kind in their assertions?
At what point do they go "I don't know"?
Do they concentrate on style or substance?
do they raise an issue with a lack of comments?
do they ask why the description of the PR is so vague?
When they get a push back, are they aggressive?
All of those are much better tests than "rewrite a thing that a library would do for you already"
No, what this shows is that the skill range for accomplished professional software devs is absolutely massive. What these companies want is to find the tail end of this very wide distribution. Leetcode interviews do a decent job at this. If you have been coding for a decade and can't do leetcode mediums with almost no prep, and hards with a moderate refresher on data structures, then you're simply not the in the right tail of the skill distribution and they don't want you. This is what so many in our industry can't accept: you're just not talented enough to earn a FAANG job.
Background:
I'm in a FAANG type company now, a YC company, 3000+ engineers. I'm a Staff SWE with 20+ years of experience (ECE degree) and make $600k+ per year. I've went through the promo cycle here (it sucks).
I can't do any leet hards and can do leet mediums after studying. Some easy's take me a couple of tries. I usually do very poorly in interview coding exercises.
** throwaway , main account is 10+ years old.
If you are lucky and join/lead/get a new project off the ground, then you'll also grow.
If you're just trying to move the dial on a normal project, then its very difficult to make any kind of headway. You are surrounded by overly enthusiastic hall monitor types who will put in more hours than you, or post more than you.
> [describes self, story completely contradicts parent post]
I'd say you're a perfect example of why "Leetcode problems are a good filter for getting the top 0.X% of talent" is false.
I mean it doesn't, because I'm at a FAANG, At a FAANG you are infantalised from the very start, sure you passed a very difficult interview where you have to balance a binary tree efficiently as possible. But you're going to use none of those skills here.
what you actually end up doing is copy/pasting some random code you found using internal code search, because the sensible way of doing it can't happen as that would involve porting a thirdparty library, and doing all the procedural work that follows.
so you hack some shit together, ship out out and hope that it doesn't break. You then decommission the nasty hack you shipped last year and claim credit for innovating. Is your product not hitting the right metrics? loosing users? doesn't matter, so long as the super boss is happy that you've hammered in the REST API for the stupid AI interface, you're not going to get fired.
In a startup/small company, if you fuck up, the whole place is going under. Need metrics? you'll need to find a small, cheap and effective system.
Here, we just record then entire world and then throw hundreds of thousands of machines at it to make a vaguly SQL interface. Don't worry about normalising your data, or packing it efficiently just make a 72 column table, and fill it full of random JSON shit. Need to alter a metrics? just add a new column.
In short, don't praise or assume that FAANGs are any good at anything other than making money. They are effectively a high budget marvel movie, Sure they have a big set, but most of it is held together with tape and labour. Look round the side and you'll see its all wood, glue and gaffer tape.
FAANGs want the top .1% of developers, they don't necessarily need them for most roles. But the point is to hire developers that you could put into any role in the company within reason and have them be successful. 99% of development work at a FAANG is pretty unexceptional and doesn't require exceptional developers. They hire for that exceptional 1%.
FAANGS want a load of loyal, naive people who are willing to work loads of overtime and not ask too many questions. Who better than posh kids from great universities who haven't quite figured out that life isn't a meritocracy yet!?
Sure they also want the top 0.1%, but they have a different interview track. Do you think all those OpenAI engineers that were going to follow Altman were asked to do leetcode?
I don't know about OpenAI specifically, but I have heard of interviews at other top ML research positions were partly based on leetcode problems.
Sure I can. By the time you get to Leetcode hard, these aren't just "can you derive the answer". The questions by design take 45+ minutes and have some weird quirk in it that is nominally related to the core concept being tested. These aren't necessarily meant to be done on the fly during an interview period.
>Not being able to see up the mountain doesn’t imply there are no climbers above you.
a better analogy is that youre on a road and you see a freeway above you. The people above you aren't "better", they are simply on another road, to another destination. But they aren't necessarily worse either. They could be on their way to a dead end job or could be a billionaire CEO.
That is to say, it's useless comparing yourself to other people you don't know. Everyone has their story.
So I'm not sure we're interpreting the same conclusion here unless you think trick questions are a mark of a "good engineer".
>You cannot argue against general rule by pointing out specialized exceptions.
Lertcode hards in interviews have been getting more common for a decade now. At what point do we stop pretending that more cats aren't losing their 4th leg and announce an epidemic?
It shouldn't be common, but it's becoming more common. I haven't seen any claims in any conversion even disagree with the notion, let alone provide hard evidence.
This sounds like you want to penalize students who studied for the exams. Or at least not reward them.
Like all interview formats, it’s a proxy for understanding if the prospect would be able to get the work done and be a good fit with the rest of the team. I’d say it’s a pretty good proxy for work ethic at crunch time as well.
If your complaint is that a normal person wouldn’t have the time to study these things in detail, why would a company want to hire someone who has external obligations?
s/exams/interview/g
External obligations like full time employment and a family?
Interviews are like exams but you don't have any clue what topics are on the test. If Leetcode was some licensed, standardized approach to getting some license to verify my ability to code myself out of a paper bag: I'd hate it, but I'd grit my teeth and study it. exams can be studied for.
But it isn't, so I can be studying leetcode questions and be hit by a dozen other topics. I don't have time to study everything, and the market right now isn't worth pinpointing specific companies unless you have a stellar reference.
>I’d say it’s a pretty good proxy for work ethic at crunch time as well.
I couldn't disagree more. Someone so prepared for interviews have the least skin in the game. Layoffs come and they didn't work > 40 hours but were otherwise excellent? oh well, get another job in a month because interviews are a breeze for them because they breathe DS&A.
>why would a company want to hire someone who has external obligations?
I'd love to know the answer here as well. Why do companies internally penalize workers who were laid off, but then try to "steal" currently happy employees but make them jump through these hoops? The logic seems backwards; interviews should be hard and depress wages for the laid off, desperate workers so you can get a desperate unicorn. But you want someone who lacks the time to study because of current job obligations to get to an offer faster. Their proof of work ethic is being employed to begin with.
Most companies are building products anyway.
We had one person who worked on [censored] 20 years ago, then was manager of [non-programmers, rest censored] - now wants to get back into coding - can this person still code? If so I want them, but if they have forgotten everything... I of course censored details for privacy reasons.
This is how I learned mathematics for example.
After you have learnt enough of the special cases (trees) you start to be able to connect them and see their relation (the forest).
I highly doubt you are sight-reading atomic bomb LC hards.
It can make the process fairer but it's not a given. You can do the classic "no dogs, no blacks no Irish," and that was a standard across pubs in England. It certainly wasn't fair, unless you were racist.
If you're committed to fairness, then a standard will help. It gives you a clear point to fix and improve things and something you can use to measure if you're achieving your stated goals. That's definitely something you can't do if you're just making things up as you go along.
And yes, I made a deliberately provocative statement. I'm obviously not saying whiteboard tests are the equivalent of segregation.
I've done tests like this for some companies. It felt a lot fairer and more closely resembling the actual work environment than live leet-code interviews, with biased interviewer(s) and a stress factor that's not a part of the actual job.
They will also be able to use ChatGPT on the job. And StackOverflow. And Google. If they know how to use tools available to solve a problem, that will benefit them on the job.
If you're testing them for what ChatGPT can already solve, then are the skills being tested worth anything, in this day and age?
Take-home LeetCode, even with cheating will still filter out a good chunk of candidates. Those who are not motivated enough or those who don't even know how to use the available help. You'll still be able to rank those who solved the task. You'll still see the produced code and be able to judge it.
Like other commenter points out, you can always follow up the take-home LeetCode. Usually, it becomes apparent really quickly if a candidate solved it on their own.
I wonder if either of the following could be cost-effective:
(a) Fly the candidate to a company office, where their compute usage could be casually monitored by an employee.
(b) Use high-quality proctoring services that are nearby to the candidate. E.g., give them 1-2 days in a coworking space, and hire a proctor to verify that thy're not egregiously using tools like ChatGPT.
Or alternatively, would it suffice to just have a long conversation with the candidate about their solution? E.g. what design trade-offs they considered, or how might they adapt their solution to certain changes in the requirements.
This is what privilege looks like. The inability to see barriers that affect others in a worse position.
Standards do not imply fairness only consistency.
You got a test that filters out people who lack the time to train for these tests. Basically, devs with a life (see family)
Is there a solution that you'd recommend?
...a standardized test? No. There are tests. They sure as heck aren't standardized.
Maybe they should be, since everyone seems to be doing the same thing.
They're more like a standardized test than a pure shoot-the-shit interview is, but still pretty far away.
Many companies have question banks that are specially designed to be fair/have some contextual relevance (ideally) to some "realistic" problem. Or at least, many of the companies I've interviewed at follow this model. I consider these coding questions to be "leetcode" style because at the end of the day they are an isolated coding problem, even though they may not be a problem from leetcode verbatim.
Companies that execute on that style of interview well are generally fairly pleasant interviews, at least in my experience. Good companies/interviewers will gauge more than just the final code to determine a hire or no hire. And a large portion of companies also have hiring ratings on a scale to make it less binary.
Question banks that are too small: leaked on eastern forums immediately, candidates show up reading answers out to you (some of the guides include guidance on when to pretend to think, I am not kidding).
The idealized version of "question banks" might work. The real one does not; you'd require employees constantly scouring forums in every language known to mankind, immediately removing anything that gets leaked. On top of that you'd probably require a competent committee overseeing all questions in the bank constantly and ensuring the lack of variance in difficulty.
Source: I interviewed at and for Goog and Pltr.
There will be bad actors regardless of the interview style, thats why companies have multiple interview types/styles/rounds to sus out a candidate, as you probably know.
If they BSed their way through a leetcode interview, then they probably won't make it past a behavioral interview where they have to go in depth on some past project. And if they BSed that as well as every other round, then hey maybe they are crafty enough to succeed at the actual job.
I think this is where our different opinions come from, while we agree on the other aspects.
In my personal experience, I have never felt that the hire/no-hire decision relied exclusively on my ability of solving the presented problem; I have passed interviews where I did not solve the LC-style problem optimally but I communicated clearly, picked up on hints, was aware of when I hit "walls" and provided working but less than ideal alternatives when I could not figure out the neat tricks.
Reading through the thread it seems that my experience is not universal, and the majority here have had less pleasant interviews, so I understand where you are coming from.
I have been rejected for one mediocre interview among many good ones. Or the other way around accepted even though I didn't perform well.
Sometimes the interviewer works with me. Sometimes against me. Sometimes a war story impresseses positively, sometimes raises suspicions.
At this point it feels like gambling.
I have also ran almost 400 interviews on the other side over the years, and to me it seems quite clear when somebody cannot write code at all. I like to think I am not biased. But who knows.
it changes immensely based on the job market. I've defintiely tanked some inerviews hard, stumbling on softball questions that shoulda been a bullet point. But I get pretty far or even gotten offers.
The last 12-18 months though? I've had interviews that felt like a dream but got zero follow up to. Been ghosted after seemingly final rounds where I spent 5+ hours on technical tests. It's not even enough to "understand the problem and communicate steps". You gotta be flawless, and you still might be cut compared to 3 years ago where a "C" performance could still land multiple offers as long as your experience made up for the quiz questions.
I dodged the .com bust because I worked for the U.S. DoD at the time.
But I got laid off for the first time during last year's "15% bloodbath".
If I compare my current job search vs. all of my job searches in the past:
(1) As parent comment said, the bar seems to be much higher. I've thought that I did really well on some interviews, only to not get an offer.
(2) Some interview processes are way more rigorous. For a DevTech role within nVidia, I had 12 interviews + 2 take-home problems. (BTW, the take home problems were incredibly fun. Well done nVidia!)
(3) I've finally accepted a job offer from a large, established tech company, and the pre-onboarding process is amazingly slow. I accepted the offer a month ago and still don't have a start date. In a better job market either (a) they'd probably work harder to be good about this stuff, or (b) I'd just take a different job because of the delay.
(4) Ghosting candidates seems common now. I'd never experienced it before now.
That's ridiculous, tho.
Did they really not have enough information on the 11th interview to know whether or not they wanted to hire you?
FWIW, the take-home problems were the most fun I've had in front of a computer for a long time.
That made up a bit for the Marathon of interviews.
What exactly are "Eastern forums"? "Eastern" what? Europe? Asia? The world?
I'd prefer to not be more specific; I chose the word Eastern on purpose.
For DEI committees reading this: I am both eastern european and asian, so I hope to be exempt from any scrutiny.
I wanted to say Eastern, as I have examples from both Asia and East Europe.
I wished to not be more specific so as to not derail the discussion into "user kidintech implied that nation X has a tendency to cheat".
Also, IMHO, blanket generalizations like "Asia" and "Eastern Europe" in such contexts can actually be more offensive than just mentioning the one country where the thing happens since you're basically painting with tar a whole sub-continent with dozens of different countries, just for the things happening in one country.
What I mean is, if by Eastern Europe, you actually mean some dodgy Russian forums, I think a lot of Eastern Europeans from Bulgaria all the way to Poland and the Baltics might feel offended of being included since we are not the same.
You are interested in Eastern Europe, so you should be familiar with olympiads; ask around any circle of ex-olympiad participants and you are bound to find something.
I am not sorry if you took offense; you're either from one of these countries and are clueless to the circles that exist next to you, or you are not from one of these countries and are trying to be offended on behalf of others.
Obviously I'm not a golfer.
If it's anything like High School/College, the sad part is these kinds of people could probably do well in interviews regardless. These answer banks are simply the difference between an A and an A+. And sadly the current market seems to only want A+ candidates. Who's fault is it really?
You mean the current FAANG market paying top dollar. I know plenty of unknown companies taking B candidates because they aren't paying top dollar (in Europe at least)
>Who's fault is it really?
The governments and central banks for devaluing the currency post-2008 with their zero interest rates, causing the value of savings and wages to plummet and the value of assets, housing and stonks to skyrocket, causing people to chase get rich quick schemes on the stock market and on the jobs market. Coupled with the VC promoting for an unsustainable growth (more like pump and dump) of so called "tech companies" who's products were not economically viable from the get go, they just survived on zero-interest money and fake promises, artificially boosting the demand for SW workers casing many young people to go into tech just to chase money, money that's now gone and so is the demand for coders.
Without this artificial demand for devs caused by zero rates and overhype in the tech industry of financially unsustainable products that were banking on skirting the local laws (AirBnB, Uber, etc), then those people chasing money would have went into finance or investment banking to chase money there instead of causing a huge backlog of candidates in tech. Just my 0.02$
in my experience: not these days. And I work in games, so "top dollar" was never in question.
>My 0.02$
That's a fair take. Tech in some ways was indeed a necessity with a huge reach, so I don't think the overhype was the difference between being a trillion dollar industry and a million dollar one. But tech would probably still be a billion dollar industry without all the factors you mentioned.
Games... well, stability was never a factor, and I knew that going in. It's a shame they are doing the exact same pump and dump schemes tech has fallen into. And it's not like layoffs were uncommon after a project finished; it's just that they are doing it purely for better looking earnings call instead of "we cannot keep funding studios anymore".
Or by "cheating" are you specifically referring to the lying cheating treasonous fraud from New York who was just found guilty of 34 felonies to cover up cheating on his wife with a porn star to inflence the election?
Why are you double-dipping on examples? If people don’t know their geography it’s meaningless, and if they do (like me) it makes you seem disingenuous.
Oh but then you bring up politics so it makes sense. Fuck this bullshit world, it’s so exhausting.
This only works if because I'm western and white I must ALWAYS be scrutinized.
This is ironic considering these companies have forced mandatory DEI seminars (which I have no problem with btw), inclusive language, #EveryoneCanCode, and so on.
But despite all this, you end up with teams and organizations that are 99% of the same X somehow. Replace X with race, school, even state from home country sometimes.
You know there are websites where people share all the interview questions to these hard interviews / referrals exclusively in their language and behind a pay or rep wall? And there are Telegram groups where international students leaks the questions or do interviews in place for one another.
It’s inevitable these types of issues arise when there’s so much at stake, ex: in just 5 years, a $200k TC advantage at a top company, becomes $1m or more with appreciation.
I just really dislike the veneer of “fairness” when there are so many problems with the process, even beyond the questions that have nothing to do with the job.
I really do wonder what those sanctimonious sermons are meant to accomplish. People who are already ideologically aligned with them won't learn anything new and may just resent it, while people who aren't aligned won't become aligned as a result of that "training".
But you're talking about it as though you expect it to have a constructive impact, resulting in irony when it doesn't. I don't see the irony because I don't expect any benefit from those struggle sessions in the first place.
Though usually the intersection of people running DEI initiatives and those people make a large set. Assimilated gay people love talking shit about how trans people make "us" look bad.
Suppose there was a SWE union and the union has to create their own reference compensation bands. Would the union again lean towards leetcode or some other standardized test? Or just use years of experience and maybe past projects?
When weighing the “effectiveness” of leetcode interviews, it’s important to weigh that thus far SWEs have failed to effectively unionize despite e.g. the past class-action no-poaches clearly showing C-suites should be paying engineers a larger share.
The reality is, fizz buzz got us 75% of the way there. It turns out when pressed, a lot of people can't write code. Yes, there's false positives, but there's also people brute focing their way through via copy & paste.
This doesn't manifest as a person who can't do any task, just as a person that's slow, delivers weird abstractions, and would take a lot of your time to get anything useful from.
But those people are also making those arguments, because, as you said, there's hundreds of thousands of dollars in it.
The failure rate of FizzBuzz has always struck me as depending on the idea that you can do a lot of programming and just never need that operator.
I only know it well because it was covered near the beginning of one of the first programming books I picked up (on Perl 5) and it stuck with me because it seemed wild to me that they had an operator for that.
So even in a language w/o a mod operator, it's not a hard problem if you understand how to solve problems with code.
But without any "real" math at all you can do it with, eg, two counters and some if statements. Or if you recognise that there's a repeating pattern you can work out that pattern manually and just write code to emit it over and over.
At least for me it would be sufficient if the person used a function like IsMultipleOf(x, m), or Remainder(x, n). This would at least make it clear what the function did even if they didn't get the exact operator.
The other thing to note is that the mod operator works differently on different languages and platforms.
Not with positive inputs, which is the domain of FizzBuzz.
Even if you don't know what "mod" means, if you have no idea know what a remainder is, and that the problem calls for it, and you can't derive the mod operator using integer addition, subtraction, multiplication, and division, then your math and problem solving skills are pretty weak, which FizzBuzz tests.
We would both test the basics as well as go through with the candidate on how they think, how they collaborate, help them out if we felt nervousness was impacting them showing their skills, and in the end got a much better grasp on how skilled they were than if we were looking at Github repos or giving DS&A trivias to solve.
I suspect quite a lot of people who are labelled as "can't code" are freezing like I did.
Do you ever have a conversation with these people as to why they think they couldn't do it?
really? they get 30 minutes + internet and they couldn't google "javascript how to get divisible by 3"? That's just bad research at that point.
However, counter point, I've had people forced on me through much of my career who just can't code for the most part, and despite being pretty reasonable about it ( it's part of the nature of the work I do ), I'm very rarely surprised by competency.
The specific failure was that I first attempted to solve by using repeated subtraction. The guy kept asking me to "solve it a different way", or saying "there is a better way to solve this". I tried using arithmetic tables, I tried using results about base10 remainders and I even tried using one of the corollaries to Fermat's little theorem to speed it up for larger inputs... every time I was told I was getting it wrong because "there was a better solution". In the end he pointed out that the only solution he would accept was use of the mod operator.
Since then I have actively kept a tally: I naturally use the mod operator an average of twice a calendar year, it has always been in personal life code when dealing with complicated geometry problems, the bit of code containing it almost always fails on some edgecase because at the point of using mod it is convoluted.
a % b = a - (b * a/b) /assuming a sane language with integer division, else cast a/b to int/
Figuring the above operation (or getting close) is when you should more or less pass, and That's a good point to show the interviewee what the operation is. That should be the point of an interview problem, to show the thought process, not that you know fancy tricks.
But alas, I was shown an XOR swap in an interview last week and spent 3 minutes deriving it on paper instead of 3 seconds saying "oh yeah, a => b and b => a" to a trick that I haven't seen since college some decade ago. The current market loves tricksters, I suppose.
And yes, the actual real world use of modulo is surprisingly sparse, despite easily imagining situations where you could use it.
``` float(9007199254740993) == int(9007199254740992) ```
If I need a "Non-Programmer Weed Out" question, I'd rather give a problem that is 1. as easy as FizzBuzz, but 2. is just 'if' statements and loops and cannot be solved by knowledge of a particular operator (or bit twiddling tricks).
This is something I can easily solve in 5-10 minutes, correctly handling every plausible corner-case.
I'm curious how common this is, statistically speaking. I'm also curious how it correlates with other things about the interviewee.
As an early 2000s integrator/analytics I learned to write code when someone watches out of interest or straight time pressure (still taxing sometimes). But most developers that I know intellectually curl up in such situation, regardless of their skill or performance levels. We worked with one very math-y low-level-skilled guy whom our clients described as “literally freezes for an hour without moving, should we pay for it?” when he did field work. He was a very strong developer otherwise.
Not that a company must hire or want these people, but the idea of writing code under uncanny pressure all day makes as much sense as that swordfish scene.
Further still, I get push back with folks citing self-prescribed medical conditions, but the same people generally display the same behaviours during the working day.
So other than contract to hire, which typically limits you to people with a job, I don't personally have a better way.
I can't speak to the statistical claim that "those people are more likely to freeze no matter the interview."
But just my personal anecdote: I do fine on take-home coding tests, but I freeze up on live-coding tests.
This is one reason I'm vexed by the allegedly common cheating in take-home coding tests. It makes employers suspicious of the testing style that I'm best at.
For north of $2m over your career, cheating probably is the smart thing to do, especially for a borderline candidate, as there's a fair amount of evidence that the prestige on your CV will make things easier going forward.
However, my problem with take homes was never that the candidate would cheat, but rather they'll probably spend way more time than the 2 hours allocated.
I'm actually less worried about the candidate doing that, than I'm worried that the interviewer bakes in a bunch of assumptions like having a machine setup to do the task, having the specific domain knowledge and experience, and then accidentally trolls the candidate with little to no avenue for feedback.
What tipped you off?
Also, I'm curious: do you think having them discuss their solution in depth would have been a good countermeasure?
Of course, this doesn't mean I've discovered 100% of the cheaters, just the obvious ones.
> Also, I'm curious: do you think having them discuss their solution in depth would have been a good countermeasure?
To some degree. But not everyone communicates as well as they code or vice versa, and then it comes to what you're trying to qualify.
For those wondering what the alternative is to leetcode: a work sample test. Take a real problem your team solved recently, distill it down to remove internal-specific knowledge, and give candidates 3x the time to solve it compared to what it takes an internal dev to solve. Bonus points for having a rubric guide the scoring.
- Input from the interviewer reflects communication ability, cultural fit, and so on of the candidate.
- Input from the blind grader reflects ability of them as a team.
- Low grades count against the interviewer as well.
Bingo; now both are stakeholders in success.
OTOH, seeing how the interviewer reacts to that situation will tell you a lot about whether you really want to work there.
If you want to earn top money, you gotta play the game. This is true in every profession, you think lawyers that want to work at top law firms have it any easier?
There's a lot of valid criticism with respect to today's tech hiring practices (especially because it doesn't only affect top paying companies), but I don't feel like "I can't get ridiculously wealthy without selling my soul" is a particularly compelling one.
Everything is a proxy because, in most scenarios, there is no way to understand effectiveness of an employee until they have worked for you for a while.
Of course there are other ways, but they all carry their own risks.
That’s why referrals, imperfect as they are, carry so much sway. In theory they reduce uncertainty and require some expenditure of political capital.
The absurd filter is just some kind of lottery. They could have a different one: at the end of the day, it's only when the person starts working that you can actually see what they are worth.
The thing with a filter like this one is that it filters out a whole category of people who may actually be good. And it reinforces itself: you hire people who are good at leetcode, who will themselves possibly be good at hiring people who are good at leetcode. Does a company of leetcoders perform better than a company made of a diversity of good engineers? Not clear.
Very much a tangent, but what do you mean by a company being "ethical"?
I have a few concerns about how that term is often used in these discussions:
(1) it's treated as a binary quality, rather than some kind of continuum, and
(2) there's rarely a recognition that different persons have different conceptions of what's ethical, let alone a justification for what the writer's preferred definition is uniquely superior.
Say you got the same salary working in a company that you found more ethical, would you go for the FAANG?
Everyone is different, but it's amazing how many excuses we can find to justify doubling/tripling our salary :-).
I don't think this is an issue of bad luck, its quite probably to get difficult interviews especially if you're applying to the popular names.
But that would dramatically increase worker mobility, for one thing. So they don’t do that.
(“No it’s because they don’t want you to be able to test prep” that makes no sense because 99% of the way to prep for these is… exactly like test prep)
higher education, even job experience is no longer considered a sufficient barrier to entry.
reminds me when the guy who wrote homebrew failed the apple interview. if i recall.
ironic because they use brew so much internally.
Well, that and how almost nobody who successfully finds employment after "grinding" leetcodes wants to remove the barriers for entry.
I think Leetcoders can't envisage a better way to assess someone than by subjecting them to the same kind of hoop-jumping you get made to do in university. They're not interviewing you for a job as there's no module on interviewing candidates on the CS curriculum, and don't have much professional experience outside of academics or software engineering. They're simulating a dissertation defence, because that's how they were assessed for their competence.
That's my charitable interpretation. If I'm being cynical, it's elitism - a way of making sure you're "one of us" (read: obnoxiously academic, Type-A personality, "logic over feelings").
Not that many, but its interesting how many are from "elite" universities. I'm in a research org for a FAANG and we often get to see all the handwringing about how we can't recruit more of people type x.
Well, if you only hire from MIT, Stanford, Oxford etc, then they are all going to look the same.
For those outside the research org, its a bit better, but its still the most uneven place I've worked.
I prepared thoroughly for all four main questions and any other plausible ones I could think of. I practiced writing solutions to ensure I was fast enough for the interview. Additionally, I pre-prepared ideal answers for each question in case I got stuck.
When the interview came, I got a total curveball: a question that was significantly harder than the usual ones. It didn't fit the round's theme (it was a DSA question, but I'd already aced the DSA round), was obscure enough not to be on LeetCode, and required writing a solver for a hard variant of a known algorithm. I panicked, copied the prompt into ChatGPT (despite being instructed not to use it), transcribed the result, and pretended I had recently studied the relevant algorithm.
I passed the round, nailed the other interviews, got the offer, and accepted. Later, I found out that interviewers are instructed to pick one of four specific questions for that round, and the one I got wasn't in the list.
I'm left wondering if the interviewer was trying to sink me or was just bored with the usual questions. The whole experience raised several questions for me:
Is it cheating if I already had pre-prepared answers for the questions they were supposed to ask? What's the difference between using pre-prepared answers and using Google or ChatGPT during the interview?
If the interview had gone according to plan, what was I actually demonstrating? My ability to use Google?
When the interviewer asked an impossibly difficult question, I would have failed if I answered it legit, even though I'm a good engineer. Failing such an unfair interview round doesn't serve the company's interests.
What is this interview process meant to demonstrate? My true value as an engineer lies in my ability to communicate clearly, think outside the box, identify and address technical tradeoffs, mentor juniors, and propose technical solutions that meet requirements while minimizing risks. Yet, I'm expected to solve a hard variant of the Traveling Salesman Problem in 45 minutes or I don't get the job? Why?
The whole process seems broken, but I'm not sure how to fix it.
IIUC, the interview deviated from the company's interview policies.
Could the answer simply be that the company has no intentions regarding the aberrant interview process that you experienced?
If Google engineers all wore pink robes and top hats while interviewing then we'd see that everywhere.
If this was just FAANG I wouldn't even mind it. Bigger compensation means bigger hoops to jump through, artificial or otherwise. At least those bigger companies come with interview guides.
But I've been dropped Leetcode mediums on some Unity game dev jobs barely paying 40/hr with minimal benefits. Game interviews already have an absurd number of topics to try and guess you'll be quizzed on. (last interview wanted low level C memory/bit manipulation question... the one before that did Unreal Engine trivia... then the one before that decided to give me vector math and ray intersection questions), adding in random string manipulation questions when you will barely use more than concatenation on the job doesn't help with studying.
Game dev is lower paying, less structured, and more involved than typical web dev jobs.
I’d really only want to get into game dev early in my career or as an independent creator. It’s a touch exploitative.
- CPU/GPU architecture
- graphics/shader programming ( I do enjoy graphics programming, but I have been asked these questions despite not applying for graphics roles)
- GPGPU
- traditional software engineering (programming patterns, architecture)
- netcode (I never applied to a network engineering position. I in fact actively avoid that part of the stack)
- general industry questions (what kinds of bugs can appear on shipping build that may not appear on debug builds?)
- a myriad of gameplay programming paradigms
- coding tests in Unity, despite applying for an unreal engine position (this has in fact happened twice now. I know Unity, but come on: can you imagine getting a javascript test for a c++ position?).
I've been asked all of these in some capacity over the years. At what point is it on me for not being some sort of omni net/graphics/engine gamedev that can answer any question on the fly, and not on the company for being paranoid about professionals studying for the skills they are clearly seeking? If someone can "study for your interview", why wouldn't you want that person? That's what they do on the job after all.
> I’d really only want to get into game dev early in my career or as an independent creator.
I'm already 8 years in, so a bit late for that. My plan is eventually to go indie, but I need a bit more time in my career before making that jump.
Despite the common advice c. 2022, getting a boring cushy tech job right now isn't very viable, so may as well stick to what I'm actually getting interviews for.
I think at least half of all programmers got into it because of video games. They know it's desirable, so they can filter almost as hard as they want.
Like, the industry is just way more jacked up than general software development. I understand your pain.
Although, if you're currently between jobs right now, you could crank out a prototype of something and see where it takes you while job hunting. Because you will always feel like you could use just a bit more time. Sometimes you just gotta do it.
I had this whole pipeline where I figured
- okay, I don't have a master's degree
- there aren't any junior graphics roles.
- I'll work in industry for 4-5 years in whatever I can while working on rendering tech, and then get a graphics programmer job this way.
- I'll be in a cushy, in-demand, highly skilled, more protected part of industry from the layoff hustle and bustle. You don't throw away engine programmers at the blink of an eye
Alas, the industry shifted around in the blink of an eye. Most studios cared less about fundamental graphics for an in-house engine as they all just changed to unreal. I did in fact get an engine programmer role, but was still first on the chopping block for layoffs. The industry tighted the heck up over the last 18 months and my side hobby projects on github barely mattered as they wanted 5+ years of graphics experience. Back in the catch 22 I was trying to avoid with all this prep.
I wouldn't say my dreams are crashing down, but it sure did show me that nothing is truly safe, no matter how experience or well liked you are. Nor even how exploited; definitely had freinds completely underpaid that put 80+ hours into the company, and they still get cut the moment RTO is announced. You have the dream capitalist candidate but can't handle not seeing them 5 days a week with butt in seat.
>if you're currently between jobs right now, you could crank out a prototype of something and see where it takes you while job hunting.
That's the weird part. Prototypes are normally a great way to stand out. But the market here is sending all kinds of weird signals. I don't know what companies want anymore, clearly not a plain ol' hard worker.
I am still getting interviews, but they are cut short in the process or am simply ghosted. Some are botched tehnical interviews, others I have no clue. So more prototypes aren't necessarily solving my personal issues.
See, this is true eventually. But it's _definitely_ not true in 45 minutes.
However, we do require a base level of competency. We give out a 40-ish minute assessment. Two multiple choice and three coding problems.
And they’re easy. One of them is essentially “John has a 5 gallon bucket and 3 gallon bucket, how many buckets does he have?”. All you have to do is rewrite the clearly labeled circuit diagram as a Boolean expression.
Not a single person this round has passed. Several have failed the simple ones.
So while grilling people on the minutiae of a language or asking them to solve the traveling salesman is not beneficial, neither is nothing.
We should be testing floors, not ceilings
For companies this is very rational, you want non-distracted employees who can work over time.
In all my years of sw dev the number of times that would have helped vs being able to communicate and manage expectations across a swath of people is like 1:1000.
with all due respect I believe this is about as far from the truth as it gets. I believe the issue is that people think they have to work at FAANGs in order to be paid REALLY well which is just nuts.
I know many people making (very) high 6 figures working at places most have never heard of. instead of looking for FAANGs people should be looking at companies that have been in business for 20+ years (if publicly-traded company the easiest "measure" would be whether you would invest your own dough into that company) and then make your self indispensable there (this is much easier than it sounds) to a point where you are more valuable to the company they company is to you. This is un-achievable at FAANG but this should be everyone's career goal. once you are worth more to a company then company is to you, compensation-wise sky is the limit :)
You know people making ~800-900k at places most people have never heard of? That is what "high 6 figures" means. I have to assume you mean more like ~$180-190k?
Not trying to brag that I have well-paid friends, though it might come across that way; just corroborating the fact that it's possible for late-career professionals to make that kind of money at non-FAANG companies.
If you can get your foot in the door it's hard to beat FAANG comp on a risk-adjusted basis and they do actually employ quite a lot of SWEs, which I think is why it's such a focal point for tech jobs even though you can get similar comp in other situations.
the entire system from the ground-up is setup to convince you that you are an individual and you need to go find a job and work like everyone else (hopefully with hefty student loan and other debts) and in our industry be treated like cattle and be subjected to some crazy interview process where you have to hack on some sh*t no one cares about.
if you start thinking about yourself as a business your goals become more like “how do I make myself profitable, how do I get to the point where another business cannot effectively operate without my services.”
is this a lottery? maybe… this is easier said than done but I would rather spend my career in this chase than be an employee somewhere where 99% of my peers don’t know my name
Why don't you list them here to illustrate your point?
That depends very much on your world view. If you get the job, it would imply you've just cost a lot of other people $100ks. That simply can't be true, because there's a small number of such jobs.
The only way this can possibly hold is if you hold that only the "best" candidate should get the job, and by "best" you mean: the one that gets most of the leetcode questions right. But then there's no us.
Edit: I'm not in favor of leetcode interviews, and I do understand that there's a bias in interviews (which won't go away when you drop the leetcode).
Leetcode is lazy, easy.
I do not know how hiring is done at megacorps; but on the few occasions I participated in tech interviews, we were looking for candidates to join the team that I was on. So, what I was interested about the candidates was how well they were prepared for the kind of work our team does; and how much I would like to work with them. I am sure this second question introduces all sorts of subjective biases; but then, hey, aren't you going to spend countless future hours with that person? At least the first concern incentivised us to make sure candidates were sufficiently competent. BTW, we didn't ask leetcode-type questions.
> but to the individual, this is costing us 100s of thousands of dollars, because there’s only a handful of companies that pay well and they all have the same interview process
I am not sure I am following the argument. If there is only a handful of companies that pay well, and if you only consider companies out of that list for employment, isn't it reasonable to predict that there will be lots of other candidates who want the same, far more than the number of available positions at those companies. How, then, should those companies select from such a pool of candidates?
If what Google search tells me is true the average SWE tenure is only a couple of years or so.
I might need a month or two of study (or more!) to be able to handle the harder Leetcode problems, but if I was expecting to have to do this every couple of years or so I'd consider taking some steps to maintain that knowledge between jobs.
That would probably be considerably less overall effort than forgetting it and having to cram for months every couple of years.
One time in a technical interview I realized a quadtree would solve the problem they were asking. I also happened to be reading SICP and practical common lisp on the plane on the way over to the interview. (specifically the sql sdl macro chapter) Unfortunately this gave me an idea. I narrated, sketched out and pseudocoded a plist based quadtree on the whiteboard and felt pretty good about myself.
I turned around to see my interviewers were not impressed. Actually they thought I was an idiot. They didnt know what a quadtree was, and they were quite upset I used "tuples".
I assured them its very efficient, and that the compiler would take care of "the tuples".
They failed me. Moral of the story don't use common lisp in a javascript interview.
The shit soup presented at a coding interview, leetcode or take home style ones are a result of wholesale cargo cult stupidity than malice.
There's no incentive at any level in BigCo. Margins are too big for top level to care. its hard to measure, middle management layer can easily fudge metrics while driving the org effectively into the ground. Lower layer that made it through the filter, sick of this grind and are busy filing tasks sipping free coffee.
The elephant in the room is that you cannot make hiring a process driven activity.
This is org level dysfunction. There's more such exhibits - DEI, promo criteria, ERGs, re-org politics, vendor selection to name a few.
Makes sense, he is older and has two daughters.
I even considered leaving the field myself.
But having experienced leetcode-style interviews on the candidate side, it's clear to me that they are no longer about figuring out and coding a solution on the spot. Interviewers expected a solution FAST, and to match that you need to have studied and learned the answer beforehand.
Yeah this is the real BS behind the tests. Good interviewers help you manage and try to find a solution. Not just "answer is binary tree"
But that "something" doesn't have to be leetcode. There are much better (and less abusive) ways of doing the same thing. All it takes is a programming problem that they've not seen before. That's it.
whether or not you succeed as an engineer generally falls down to whether or not you know how to code (pretty much anyone is good enough at this for most jobs), can you communicate effectively (covered during the behavioral portions of interviews), and do you work hard.
honestly i think it sucks, but in comparison we're pretty lucky. i think a lot of people in this world would be happy to grind a few hours a day for a couple of months to get paid 250k+ base
Regarding “cheating” with an LLM, there is not unanimous consensus on that either. I’ve been in technical interviews where the collaborative coding tool had an LLM built right in, and I was encouraged to use it. If thats what I’m going to do at the company then demonstrating problem solving that way is useful.
> I don’t really have a solution to this problem, I just know it’s a problem.
Yeah, exactly. You don't have a better answer and nobody else does either.
Everyone knows how pointless it is. But it’s become more like paying the tax.
For the record, I typically give away the a-ha moment early on and see how the candidate is able to take the suggestion and work through a problem with me.
The solution is simple: refuse to do those interviews. I have, and I wish more engineers would. It's one of my screening questions: do you do whiteboard style quiz/brainteaser interviews? If the answer is affirmative, I politely (and sometimes impolitely) decline.
I've written a book, edited another, contributed to plenty of open source, including Golang, started a few companies, and been around the block. I'm more than happy with a take-home or with going over code I've previously written, but doing leetcode in a "live coding session"? I'm good, thanks.
For FAANG and other huge companies, it's probably a decent heuristic (though I doubt even that), but smaller startups or non-tech companies doing this nonsense is a clear red flag. I mean, there's entire books dedicated to gaming the software engineering interview. The signal-to-noise ratio is probably so terrible, it's essentially just professional hazing at this point.
We've been going in circles for 20 years.
I am starting a company, and I need smart people; I do not care how good they are at programming language X, or technology T - all the skills my employees will need can be learned on the job.
I want to optimize the time it takes for an arbitrary hire to become independent, to have learned the basics, and to make meaningful contributions. They would write code at most 20% of the time, and the job has many other nuances.
In your opinion, what would a process that you wouldn't refuse look like? Would a learning interview (I present something that you have not seen before, act as an oracle, provide docs, evaluate how quickly you pick up concepts) be so insulting?
So you are looking for somebody in top 5% cognitive ability and top 5% independence which is what pretty much every other employer wants.
My advice is if you can pay competitive FAANG salaries (or compensate lower salaries by prestige/saving humanity/bringing people to Mars), just do what everyone else is doing - Leetcode, behavior interview and architecture interview. Innovate on something else, standard interview process will give you smart motivated folks.
If you can’t pay those salaries, you need to be creative and decide what compromises you are ready to make and tailor interview process to them. No one size fits all process here. You are essentially looking for people who failed in standard process while still posses qualities you need. Expect to spend more time on hiring and have more bad hires.
I was genuinely wondering how OP preferred to be approached vis-a-vis this sort of assessment, since they suggested that they would walk out on conventional approaches. I mentioned cognitive ability and ability to learn as I feel those are harder to extrapolate from one's existing publications/contributions/take-home assessments, compared to in-person discussion(s).
Weird requirement, why? Even places like RenTech don't specifically look for the top <1% cognitive ability (and they quite literally print money).
There is no way more than 1 out of 100 arbitrarily sampled humans can win Putnam or get tenure in pure math/physics/etc. at an Ivy.
You're being very nebulous and without knowing exactly what you're hiring for, it's hard for me to recommend a process.
Gonna <press X to doubt> on this one. I mean I guess anecdotes are cool, but yours goes against basically any editor/publisher due diligence, not to mention the financial incentive of writing a book in the first place (the book actually being good), but alright.
1. Do you work hard (i.e a person who studies many problems, memorizes solutions, and regurgitates then)
2. Are you smart (i.e a person who doesn't do any leetcode but knows the fundamentals and can sythesize past knowledge well enough to answer a new question)
Now in a vaccuum, my takeaway from someone who doesn't pass a leetcode question is that they are more likely to not be either of those (given no other information about them). In my opinion, solving at least one leetcode easy - medium question (maybe with some direction from the interviewer) should be the standard for any role where you need to code.
And realistically, companies index less on leetcode skill as you become more senior or apply to more niche roles. And plenty of companies out there have bug squash rounds, system design, behaviorals, take homes, trivia, background specific questions, etc. Leetcode is usually not the be all end all for hires above entry level. Leetcode is not meant to magically select for the best engineers. Its simply another signal for the hiring committee to use.
Leetcode questions can by definition never come close to real life problems - they are simply out of context. Working on a project even a few weeks will grant a smart person any insights he might actually need. The kind of out of the blue puzzle problem a leetcode presents is more akin to crosswords in old leisure magazines.
This isn't working hard. And you are competing with the people that are working smart. Such as people having a separate computer/screen to either look up the answer to a problem or ask an LLM to regurgitate some code you can transcribe. It selects for people that are dishonest.
> 2. Are you smart (i.e a person who doesn't do any leetcode but knows the fundamentals and can sythesize past knowledge well enough to answer a new question)
See above. A lot of the problems have nothing to do with real world coding scenarios. Do I need to pull out memoization and know how to work with grammar rules off the top of my head? Or the functionality of quad trees and how it applies to partitioning? No; I've literally never used them in my career and likely never will. What I do need as a software engineer is understanding the high level concepts and when to apply them and refresh myself on the use case when and if appropriate.
It leaves out looking at the engineering skills it takes to put something past the finish line on budget.
It's at best, an adjacent sideshow and at worse a proxy for ageism to preference people closer to 19 and less capable at getting quality software out the door.
And maybe that's why the industry has such a damn hard time getting quality software out the door.
Maybe.
Here's a far better idea. Give them a broken build or crashed code or some other failure and ask them to fix it. If they can navigate around a system, diagnose a defect, find the cause, offer a patch, and match the coding/commenting style etc, that's an actual player you need, not someone who can still solve an algo question from their CS101B midterm exam at 37.
This version can go deep as well. You can have separate branches and an issue tracker, maybe with the conversation of the bug already going. Maybe the comments have a patchfile that needs to be manually applied. Maybe the patch has a new bug in it ... Perhaps it's a red herring and addressing a similar but unrelated issue. Maybe it's a closed bug and a regression and nobody wrote a unit test for it. Maybe the unit test is there and it's up to the candidate to discover it or the unit test is out of date, buggy and useless.
This is the real stuff.
Get them to communicate after. Saying something like "the unit tests were broken" versus "I was unable to understand how to get them to work right so I moved on" ... That's a huge difference in attitude - only one of them is a team player.
I've known people who program things like jet engines and stuff that goes into space. They know their stuff, are super methodical and take their damn time to do it well.
Quality programming is a rarely practiced time consuming art and you'd be lucky if you have 10% of your workforce doing it. Heck, you'd be lucky if you have 1%. Those people save the ship.
I'm not one of them. When you meet them, they're like a different kind of human.
If you timebox them to 30 minutes, the people who write mission-critical quality code will fail to get it in on time and be filtered out.
These are the people who write those 10,000 word articles that appear on HN with carefully constructed schematics, equations, and code that you struggle to understand and bookmark to study later.
On the other hand, if you give people 5 hours, I assure you, the morons will still obviously be morons and you give the "True" Engineers who take their craft dead seriously a chance to shine.
I'm sorry but it's incredible hubris to think you'll come up with say Dijkstra's algorithm during the interview from the first principles. The only reason you can do it is you studied it before.
Chance of someone
- seeing it before: 10%
- being a liar: 5%
- being the next Claude Shannon: 0.000001%
So someone coming up with the right answer and claiming they've seen it is useless.
Right answer and claiming they hadn't seen it is basically a guaranteed liar.
You can use it to filter out the liars I guess. But you'll also filter out your Claude Shannons as well.
Either way, it's a dumb test unless you're genuinely getting a steady stream of Princeton and Cambridge magna cum laudes waltzing through the door and can reliable bump the last two priors by a few decimal points.
Most people applying already have a job. And if they work hard its actually quite uncommon for them to have ample energy to do leetcode questions on the side. So you might get the opposite - you'll get candidates who aren't great performers in their current roles but used their extra energy to grind leetcode.
Even worse: it misses out on the hire people with no formal training too busy to waste time on skills with little use outside of the interview (in the real world, they'd use a library).
when does this start? I'm over 8 years in and I'm tired of being told to make a dang NxM Bingo solver in 30 minutes.
>Leetcode is not meant to magically select for the best engineers.
I'd agree with you in 2022. in 2024, I'm being way more filtered out for questions that I get right but didn't solve fast enough, or with zero hints. I was never a leetcode savant (too many other things to study for, and LC is maybe 20% of my interviews) but I don't feel that much worse than 2 years ago. Companies just have sky high expectations and are on a bargain bin for 15+ year roles paying 5 year salaries.
https://github.com/emilybache/GildedRose-Refactoring-Kata
in their favorite language. Has been a very effective filter
In any interview, I kind of presume that the interviewer has a "perfect answer" in mind and is playing some game where I basically need to guess what number they're thinking of.
I'd make it clear that anything is ok. Just go some direction
And the problem can be scaled to any complexity. Distributed/Design/Data Storage
If they give a compelling enough set of answers, I'd just skip the rest of the exercise and be done.
After learning how to do anything, the next 20 years is about learning when not doing anything is the right move.
When I make interviews I also come with the code, normally a few line snippet and have the person tell me what it does or whether there might be a bug and talk about it.
Reviews are also great. They take away the show-cooking aspect where many people feel nervous and makes it more conversational. It is also very easy to evaluate as well because everybody debugs code at some point.
I personally dislike even fizzbuzz because there are always clever ways to go around it that are totally irrelevant.
I was once asked to implement sort a list in python, so I just took the list and passed them to sort. But no the interviewer wanted me to implement a specific type of sort…this was for a distribution maintainer job and I almost failed it because of that. Fortunately the other stages were actually technology related and I aced them.
The (best) seniors just use a library. The good candidates roll their own, but can explain the tradeoffs. The juniors get the test to pass and stop.
What? Do you have something against zip codes and email addresses?
I like to ask something to do with strings - English sentences. I tell them it's intractable in the time we have and that I really want to see how they whitebox-test the code they write, so take ~10m timebox to write the code, then write some tests - at least one test that finds a bug.
I too used to hate leetcode questions, but after conducting 100s of interviews as an interviewer, leetcode style questions are the best way to test intelligence and coding skills ive found in general.
However, leetcoding interviews must be paired with system design and topgrading in order to have balanced interviews. Leetcoding alone is incomplete but i find it essential.
I disagree. It’s supposed to be that, but the reality of the interview is that there isn’t any time to problem solve so you need to learn the problems and just regurgitate the solution.
That doesn’t put you in the top 0.1%, it ensures you’ve spent dozens of hours practicing questions.
The side effect of this though, is that it's an extremely painful process for the applicant. Even if you're a great software engineer, you're going to freeze up, miss something, have a bad day, and fail a few interviews. Now for most software engineers failure is an unusual and harrowing experience, so they react really badly to it. Especially ina scenario where it's difficult to blame anyone but yourself. But just don't! It's fine! just move on, there's plenty of jobs and the interview process is a blunt tool, not a final evaluation of your worth in life.
Totally, but leetcode wont filter them out.
Incompetent is fairly easy to navigate around, pathological bastardness is impossible to work past. I know that some companies use psychometric tests, but they are not based in anything remotely scientific, but are also super easy to game.
Coding challenges won’t filter every pathological bad hire, but coding challenges do catch a lot of bad hires.
LeetCode shouldn’t be the entirety of the interview. It’s just part of it.
leetcode doesn't catch bad hires, as someone whos done a lot of interviews, we could easily catch "bad" coders with a very simple whiteboard test or a "look at this code and tell me what you think"
but people being bad at code aren't the company killers, its the toxic people who are "good" at coding that do. They can be very good at these kind of tests, but are absolute shites to work with.
I think we actually agree more that I let on, and you are right that coding tests should only be a part of the interview process.
But in fact even in at-will situations, firing aversion seems to be a real and expensive issue that all kinds of managers in organizations face [1] [2].
So, in one sense I agree, bad hires are existential threats "only" because we don't fire them fast enough. On the other hand, have you ever had to fire someone yourself? Try it. It's heart-wrenching, even to a heartless utility monster like yours truly.
[1]: https://www.econlib.org/archives/2012/06/firing_aversion_2.h...
[2]: https://www.econlib.org/firing-and-the-left/
- https://ultracode.ai
- https://interviewsolver.com
Well worth it. Highly recommend these tools to play the dumb interviewing game.
And because they ask pretty irrelevant questions, the 6 weeker will do better then the recent Stanford PhD who has been studying a specific problem for 5 years or the multimillionaire who spent the past 10 years building and selling successful software companies.
Those people will fail and your 6 weeker who's never heard of cmake, diff or gdb will pass.
When I see there's leetcode as the hiring process I presume the place is full of bozos because the process actually strongly preferences them.
Either it wasn't a true "zero" but rather they were really good e.g. at advanced math or physics and managed to see some parallels, or it has to be a genius. I'd be very interested in hiring such a person.
The code to solve basically all of them look nearly identical after you classify it and the questions asked (what's the big O of this) are also all identical.
It's an extremely narrow and teachable class of problems.
There's outliers for sure but the 80th percentile are basically just the same damn things slightly rephrased. It's extremely gameable and doesn't test the kind of breadth say, fixing a crashing program would show.
Here's a list of 16 https://github.com/Chanda-Abdul/Several-Coding-Patterns-for-... ... That's the 99th percentile.
If you think someone passing that means they can fix the conda bug in your k8s cluster or whatever real problem you have, dream on. It'd be like hiring a cook by quizzing them on the names of kitchen utensils.
The other way round (people who code, but can't leetcode) I agree it's an issue. Unfortunately as noted here in comments, cost of a bad hire is much much worse than cost of a no-hire, so it's the price I'm willing to pay.
Excellent! That’s a something, a something you can measure. Back when I was involved in hiring, it was definitely better than not having that (at junior levels).
However, the idea that “IQ is all you need” is so obviously false that I think you need to reconsider. Bill Gates famously, (infamously?), hired for IQ. Microsoft used puzzles as their IQ proxy, and I think leetcode is just a modern variant. Possibly a worse version, as training has a bigger effect.
Bill Gates later would claim that he overvalued intelligence. I would say that intelligence is necessary but insufficient. We would get further, faster, with a quick IQ test and then dig into the candidates experience, communication ability and personality. No need to study!
The FAANG companies all have a line a mile long of candidates willing to put up with any amount of BS for a chance to work for the name and $$. If you aren’t such a company, trying to hire that game changing person for $$’s you can afford to pay is going to require a different approach.
I think the mass trained at bootcamp is an exaggeration. The best bootcampers are those that use it as an extended education from learning the traditional ways. I'm sure most people here complaining about leetcode can indeed spend 6 weeks there and be interview ready. But few have that time nor spare cash lying around for such a venture. And why should we need to spend thousands just to ace a quiz like it's some SAT?
>cost of a bad hire is much much worse than cost of a no-hire, so it's the price I'm willing to pay.
At this point I wonder. Hearing way more stories about "survivors" from layoffs burning out from jobs that keep promising that they'll hire more staff to help out with. When in reality they have a hiring freeze or are trying to offshore the work that ends up needing more time to be fixed.
The stalls in a no-hire aren't just costing money in the lack of productivity, it's costing currently working employees who are doing 2,4, 8+ jobs without any meaningful increase in role/salary. It's not sustainable. You're draining a little bit of blood from a rock, but the rock is clearly starting to crumble.
If I genuinely wanted a job again all I'd do is go to one of those drilling game websites for a few nights and then say whatever the interviewer is looking for.
Any test I can fail one day and pass 3 days later that is ostensibly supposed to assess whether I have skills that take 15 years to master is total bullshit.
The initial stated problem was different: false-positives, bad engineers practicing for 6 weeks will pass the test. That I seriously doubt is a real thing. If they managed to learn that fast, they must be good. They could be junior-good -- that's perfectly fine, you need to hire juniors too. You don't judge seniority solely based on leetcode results.
I could be wrong, but I think dynamic-programming problems belong in that list.
Recently I failed the first round of a fintech analyst position because I didn't complete the 9 undergrad (or high-school, in certain talented schools) math problems (I have a math PhD) in the 1 hour time frame. In retrospect, I should've cheated using computer/online assistance, and my true failure was my unwillingness to cheat. Not because fintech analysts are cheaters, but because the testing process is nonsense, so recognizing that, one should cheat it (indeed, an analyst mindset!) I'd rather be jobless than to adopt that mentality though. (and in fact I am, the only job I was able to find was a minimum wage service-type job, it can't even pay for daycare, I'd rather raise my child on my own while wife works thank God.)
We have high-interest rates and global crises resulting in companies not hiring while simultaneously governments give orders to news media to tout the strong economy. Ineffective hiring practices will always exist, but people don't complain when jobs are available; you only see this type of discussion online when things have gone sideways.
Integrity is really tested when it costs you something. I'd call this a true win.
(I recognize there are dire situations where cheating might be warranted. I hope you're not there yet.)
I do work in a team where majority was hired by puzzles of sorts. All of them are smart. They are not software engineers and it shows massively.
Coming from Real Engineering and making 2x more programming, I lol at people who say there is any Engineering in software.
We are programmers dealing with layers upon layers of abstraction. Knowing to optimize for time by using vectors is more of an art, than a science.
I did do safety critical C code and Assembly which are probably my hole in the whole 'its not engineering'. But javascript/python/backend work: programming.
It's just a job title, arguably one without the protections or standards afforded to chartered engineers, and with an incredibly low barrier to entry. Pretty much the reason we have to do this ridiculous dance with tech tests, LeetCode, and 12 stage interview processes.
Exactly my point.
The strong consensus was that software is an engineering discipline, and the nitpicking you can have on any particular topic that is/isn't engineering-y can be applied across the board.
But this conversation is totally moot IMO because if you have a bachelors in computer science or software engineering, and a masters in the same discipline, and several years experience, you can apply and become the same chartered engineer from the same association that does "real" engineering.
I see it as a difference in time and stakes.
We’ve built bridges and boats for thousands of years, where the outcome of failure is people die.
This has a lot to do with why it’s easy to estimate the time and cost to build a house, but it’s a rare shaman who can consistently estimate software projects well.
Once we’ve built software for even a few hundred years, we’ll probably have it pretty dialed in, too.
We write software that has killed people due to bugs too, such as Therac-25 releasing too much beta radiation killing patients.
> This has a lot to do with why it’s easy to estimate the time and cost to build a house
My parents and in-laws are property developers and I've also been involved in a build; not one was on time or budget. Some are really off budget, beyond the scale software projects that mess up are. In China they've had projects run into the billions and be decades late.
I do understand your points, but this is what I mean when I say you can nitpick any issue with software not being Real Engineering and apply it to other engineering categories too.
But yeah we may have a very different process in 50 or 100 years. I'm only 10 years into my career and it's already pretty different to when I started!
One related problem I see is that Computer Science tends to be under-estimated by other engineering fields. A ton of software engineers did not study software. They learned to program as part of their engineering studies, and they believe that they learned both Computer Science and their engineering topic at the same time.
I may be biased, but I see fewer computer scientists pretend to be mechanical engineers because they once tuned a PID.
If Bachelor of Science computer science grads were that much more extraordinary, they wouldnt be working side by side with non-degreed programmers.
Chem Engineer here, I wrote random forest once. My coworker with a CS degree, all front end work.
There definitely is. Safety/mission critical code is definately a more rigorous, formal process where you don't just download a library for padding a string (you probably wouldn't use built-in strings anyway on this kind of code).
But I agree that the way most coding works (i.e. move fast and fail often) is completely antithetical to how engineers should work. But there's infinite moneys and almost zero accountability, so who's gonna stop them?
>Knowing to optimize for time by using vectors is more of an art, than a science.
that's definitely why I fully agree with calling programmers creatives over engineers. once you get past dead simple problems, no two programmers are going to approach a prompt the exact same way. performance, scalabiliy, documenting, version control friendliness, even ephemeral stuff like code style. coders very much are artists who happen to know a lot of math.
It is a downgrade in my opinion.
Most programming work needs common sense a lot more than it needs theoretical algorithmic knowledge.
And almost nobody values good documentation, especially as a competetive advantage.
FWIW I think LeetCode problems can make sense for some businesses, but most not.
However, I have a real question as a follow-up: Have you been able to profit from this skill? If true, you should be able to crush any "whiteboard" programming interview. If I had it, I would move every 2-4 years and get large pay rises. You should definitely be able to pass FAANG interviews with that skill.
But it's entirely pointless.
I've seen this reasoning, but I'm wondering if this is actually true, especially if you also believe the other myth in hiring: most people applying are not qualified for the job. Since most hiring actually requires you to fill the position in the end (in other words, you can't leave it open indefinitely if no one qualifies as eventually the pressure to hire mount up), every false negatives does increase the odd of you hiring a false positives as well. Just take the example at the extreme, if there is only ever one person on Earth qualified for your job and you got a false negatives on him, there is no other way but getting an idiot in the job.
We would work together at the whiteboard, with me teaching them the basics of our application domain and our internal distributed computation framework. Then there were a series of 5 increasingly difficult iterations we'd work through. First, do a simple calculation. Now there's a need for a more difficult computation. And an even harder one. Then explain that it needs to run over billions of log lines, and it's I/O bound, so let's make it faster. Finally, a problem that requires them to probe me deeply about the nature of the data and only needs a hand-wavy answer (calculate a median in a single pass with limited memory - not too bad once you extract from me that the samples are all integers between 0 and 500). To the candidate, this looked like iteration driven by evolving requirements, not 5 interview questions.
I mostly let them drive, but I jump in now and then to keep things on pace. And I would try to answer any question the same way I would as a teammate / mentor.
They don't need to get every problem right. The increasing difficulty is designed so that almost everyone gets something right and something wrong (< 1% gets all the way through the last one). That's an opportunity for me to give supportive feedback to see how they respond. The best responses to feedback yielded something like "oh cool, I can also use that over here to simplify the code". The worst was someone insisting that you can flip the order of addition and division without changing the answer (i.e., order of operations doesn't matter), me saying "I think it does matter, try doing 1/2 + 1/4 in different orders", and them saying "no I have a math degree, I know what I'm talking about."
What they do need to do is be able to solve problems collaboratively, extract learning from an eager mentor, and extrapolate from those learnings to solve problems. The best candidates start finding trade-offs across the application domain / tech tooling boundary that I haven't even brought up.
One of the downsides is that it takes a while to calibrate as an interviewer, and that calibration is not entirely transferrable across interviewers, as candidates will respond differently to different interviewers. I did this interview thousands of times (my record was 14 times in one day, and frequently 20+ times a week - we were scaling an org from ~50 engineers to over 1,000). I never stopped learning about signals I was seeing, but I was mostly calibrated after a few dozen candidates. Leet code calibration is much more transferrable, but what use is that if it's not measuring what matters?
One thing that I think leet code interviews get very wrong is that interviews involve two sides trying to evaluate each other. Wearing people down with leet code is not a good pitch for your work environment. Many candidates told me at the end of my interview that they were shocked at how much fun/educational an interview could be, and many hires said that this interview style was the reason they accepted our offer over more lucrative offers, because they felt they would learn more and have more fun with us.
Someone who does well on leetcode is usually easy to teach so they will become good/great in a year or two, but if you don't have that time then go for leetcode+experience just like everyone else is.
Depends where you’re looking in the hiring pipeline. If you’re looking at raw applications submitted to a publicly listed job posting, I can say without reservation that the majority of candidates are not qualified. The pre-screening candidate pool is extremely bad on average, especially for remote job listings. Some people spend all day clicking the apply button on every job listing they see.
Your post constrains another misunderstanding about the hiring process: The goal isn’t to hire the first person through the system who is good enough to pass some arbitrary threshold. The goal is collect a number of candidates and hire the best one. People hate this reality, but it’s true.
Or rather: the best fit. You don't need a Turing Award winner to clean up your website.
I definitely do not have any misunderstanding here, and my description of hiring is the same as what you wrote, just in different phrasing: after a period of time, the company has to pick a candidate, presumably the best one by their interview signals.
I am disputing the standard claim of "it's better to have false negatives than false positives" by pointing out how that mindset does not work: since you have to choose someone anyway, the whole "avoiding false positives" might not be a thing at all
As someone who’s done a lot of interviews, this is very true for SWE roles.
I've never been through the "regular" IT hiring process myself, but I've interviewed several candidates for an Android developer position. I liked asking questions and looking at how the person thinks. I didn't expect exact correct answers for every question, I just wanted to see whether they have a fundamental understanding of Android. After that, they had to build a small app at home (not my initiative) to show their skills.
There are better questions where I do think google should be allowed but they aren't LC-style. Some examples are building APIs or operating on files
“Trick” questions are dumb, but you want to get an idea of if the candidate knows their stuff (and let the candidate know what we’re like — interviewing is a sales process in both directions). If someone asks a question like “is it ok to modify the argument” (or says “I’ll assume I can’t modify the argument”) that’s great (or says “I can’t remember if strlen includes the 0 byte so I’ll add 1 — normally I’d look it up”) - again, no trick questions.
When I say “simple” it’s something like atoi or “shuffle a deck of cards”: basic, but not 100% trivial like fizzbuzz.
On of our best hires was a guy who made a fundamental mistake — when we asked him if it world work as intended he said “I’m a fucking idiot” and fixed it. You’re not supposed to talk like that in an interview I supposed but we read it as a strong positive signal.
We had one candidate who insisted that parsing a string and returning an integer was an unreasonable question because there was already a library function for it. “But what if you have an embedded system and no library?” (an actual situation for us, for part of our system). He was adamant — honestly that was a failure of the phone screen: had it been caught it would have saved us, and him, from wasting time.
It seems that “able to code a simple problem well” is a far less ridiculous proxy for good software engineer than “able to code a hard problem at all”. Much to my surprise.
The software industry is notorious for underestimating overall complexity factors, understanding uncertainty, and time to manage them and interviews seem to be no different. So “simple” problems are often good (I agree with you) because they’re actually not so simple for most people when you factor everything in, they’re often reasonably challenging under the set of circumstances.
Meanwhile genuinely difficult problems (in any arbitrary setting and longer timelines) require a whole lot of rote training and reusing little clever idiosyncratic techniques that may or may not be generally useful, so you very often end up with a question lottery on whether you already know how to address the problem in question or not correctly vs having some problem where a reasonable solution and expectations around that can be derived in the time (and environment) it’s taking place. Lots of nonsense.
Truth is, I'm never looking for "the best", I'm looking for "good enough to do the job competently". Because we can hem and haw over what is best for hours. And we could disagree with what qualities in what proportion make someone "the best". But we can all generally agree with "could do the job". As I've mentioned elsewhere, we're hiring right now and we have the duality of candidates in our pool.
There's one guy who is just a hard no from at least two people in the group. And that translates into a no from the third member simply because we're so adamant that he's not the guy. And there's another who we're all feeling really solid on. Not "the best", but solid floor and we feel they'll be able to learn.
But being able to weed out people who can't even perform the basics despite having a decent resume is invaluable.
When I worked for more medium sized companies, I'd get software engineering candidates who obviously did not even remotely know how to code. Imagine the worst possible software candidate, and then lower your expectation even further. One couldn't even tell me what an "int" was. I think a single "for loop on a whiteboard" question would filter out at least 50% of candidates.
"You can lead a horse to water but you can't make it drink."
It says barely anything about the universities and everything about the student. If a student just wants to just have a perfunctory gargle at the fountain of knowledge instead of drinking as deeply as they can, that's on them.
Almost everyone embellishes their resume. Some do it to an extreme. The first stage should be weeding these people out, and any hiring manager learns this lesson very quickly.
Story: I learned this as a hiring manager early on. I was the resume screener, first contact and then last contact. Everyone in between was my team and other adjacent teams. I failed to ask the right questions in a phone screen and when we brought this guy on-site he failed very bad. Even worse, he broke down begging for for another problem he could solve.
It was bad for everyone and we were a very accommodating group that gives many tries. We weren't looking for perfect answers to one question.
Though I still do start with a short fizzbuzz-level question because you would be surprised how many people fail that and it makes it way less awkward if you start with atoi and they can't write a loop.
If they fail fizzbuzz I generally just turn the interview into a chat or give them easy questions to practice, rather than say "yeah no thanks", because I feel like the latter is a bit mean, and everyone can benefit from interview practice.
It’s kind to politely ramp the interview down in that case but I’ve learned to then cut off any later interviews. It’s a waste of the company’s time and the candidate’s too.
If we were going to have lunch the recruiter takes them to lunch anyway, unless the (now former) candidate doesn’t want to.
Trying to debate the interview format or reject interview questions is one of my hard-stop rejection triggers in interviews now.
The couple times I’ve been part of hiring pipelines where candidates argued about the interview itself and then got hired anyway, they became extremely difficult employees within the company. Arguing about interview questions turned into arguing about every other ticket, code reviews, and every architecture choice.
I’ve had the same experience where we take an actual problem we solved in our codebase and turned it into a small interview problem. Same thing: Some candidates will try to debate the relevance or try to wriggle out of it.
if the interviewee pushes back and find they don't like the answer, then it's a good signal to both parties.
Sometimes you just have to "document your objection, but do it anyway."
Sometimes we want harmony and not drama.
Knowing how to push forward without being abrasive is a more valuable trait.
Fwiw this is surprisingly subtle and non-trivial to get right. A naive "random swap" approach does not produce uniform permutations. Sorting based on a random comparison function can likewise produce biased results depending on the implementation of the sort. Tagging each element with a random number and then sorting based on the tag works (or so I think?) but you need to deal with the case where tags collide. Fisher-yates shuffle is bulletproof but I would not trust myself to write it from memory lest you end up writing Sattolo's algorithm instead.
At the end of the day the question is: “could this person do the job and could they work with others in the process?”
this would make sense if they filled these positions in 2,3 weeks tops. But that's the insane part; I hear of interviews going on for 6-7 stages now, lasting almost 6-10 weeks. At that point I feel you are no longer avoiding false positives, but simply hiring the most desperate programmers willing to jump the hoops. The best candidates will get an offer mid interview at that rate.
>But just don't! It's fine! just move on,
I've been "just moving on" for some 9 months now. There's only so much morale in a tank, no matter how BS you know the whole process is. I just wanna move on in life, not be bogged down on how many permutations you can color an N x M grid with.
Best tests are simple and common computer science problems in algorithms and data structures with no tractable solutions. There is no facile answer, you truly have to reason from first principles for any particular scenario because there is no alternative.
I have just went through one interview that have 4 rounds.
1. half hour talk 2. 2 hours coding for several crud api. 3. system design project ( I spend 2 days on it. they told I have about a week on it) 4. another half hours with CTO.
and then got rejected. (The interviewers does not bother to send a reject letter. I asked them they told me the position is closed).
How generous. One job I asked for an update a few days after my assumedly last interview, 1 week afterwards, and then 3 week afterwards (right before the holidays)... even by the last response the recruiter simply said "we are still evaluating the best project to put you on" or something to that effect.
That was 8 months ago now. Going through 5 rounds of interviews into a rejection sucks, but at least be straight with me.
It's been proven many times that the severity of an initiation ceremony significantly boosts the commitment of those admitted to the group.
What is a better capitalistic company hiring programmers:
>Lots of smart similar people who believe in correct answers
>Lots of random skilled people
If you said the latter, you are idealistic and your opinion genuinely doesnt matter. Nature will replace you with someone smarter and realistic.
Citation needed