Hacker News Reader

Comment by hinoki | Hacker News Reader

I work for a FAANG company in a senior position and I did fail FAANG interviews w/o advanced prep.

The second try I practiced.

I YOLO’d the Google interview two times. First time I failed. Second time I somehow passed. It can be done. Might have gotten a higher level with studying but I don’t think I would have ever studied.

Off-topic: If YOLO and you want to join Google, wouldn't that be motivation to study?

Unless you view it as not worth the effort.

I've done two onsites with Google in the past essentially YOLOing it (I only study my interview failures because I view otherwise as an inefficient use of my time) - first time did terrible, second time almost passed if I didn't completely bomb my very last session. The second time ended up not really mattering because two different teams in two different orgs for my current non-Google FAANG wanted to hire me after onsites done on back to back days (side note: that was almost 15 hours of interviewing in two consecutive days - that's a lot of time, I only was able to do it because I was funemployed at the time).

I actually appreciate it very much if a candidate didn't study & focus more on giving the best answers to their capability when I interview them - the questions I give them are usually questions that no amount of studying would have prepared them for, so already taking the mindset of trying to respond thoughtfully & earnestly to problems & situations that change on a whim puts them a step ahead.

I studied for the interview I wanted (by being a thoughtful software engineer in my day job) and not for the interview they offered (which would require me to either be a professional leetcoder or some algo/performance expert). If they didn’t want a good software engineer then they’d have to pass on me. I made it clear in every session that I was thinking through the problem, asked good questions, and when there were aspects I could write concrete code for I did. If you score by solution competence and performance I think I aced 2 of the coding sessions and did pretty mediocre in the other 2. My interviewers must have been willing to go a bit off of the default mode of operation as I managed to get an offer.

I don’t know if interview performance has anything to do with negotiating power, but I was able to get damn near the highest possible total compensation my level allows for without a competing offer.

you already studied in college

Is YOLO’d when careless and impulsive behavior results in tragic and unforeseen circumstances?

I'm a former FB eng and a Xoogler and have been an interviewer at both places. The recruiter sets the level for your interviews. There's an option for the interviewer to recommend you for a higher level, but there's no incentive for the interviewer to do so and is almost never used. Getting higher rating from the interviews will only make the hiring committee's decision to hire you easier, and almost never affects your levelling.

In my circles at GitHub and Microsoft, the interviewers absolutely had influence on leveling.

After interviewing people for a while, I've learned how to suss out what people gain from practicing vs actual engineering instinct

I've passed people who "didn't meet the bar" because I could tell they just didn't practice, but exhibited 4 stars on every "Ultra Instinct" signal. Programming speed isn't important, correctness & habits are what are important. + or - 10 minutes to finish a problem doesn't really matter in the daily job

can you say more about how you detect instinct?

Sure :)

- People will catch edge cases before they occur

- People will realize that they have an edge case, and write a failing test/trigger the failure before they fix it, and explain it to me

- They come up with an approach pretty much instantly, and their design is correct after thinking about it a bit more

- Their coding style is very functional & composable (idk why I find this correlates with success, but it does)

- They test all dimensions of variability in their inputs

- They focus on what matters

- Forward progress doesn't stop

Things that don't correlate with instinct

- Speed of implementation (within reason). Different people can type at different speeds, and people think at different speeds

- Having to google APIs

I've had two memorable interviews where timed coding was part of the interview, and I wowed the management team but did not get a call back due to taking about 10% too long on the coding. This is a good thing, in hindsight, considering the eventual fates of those businesses that hired like that.

It's counter-intuitive not to test for coding at an interview for a developer. But you just can't learn what you need to know, as a hiring manager, from a pass/fail timed test. This greatly informs my hiring process now.

I YOLO'd my Google interview. I didn't even expect to make it to the on site. But I did, then I bombed.

I YOLO’d FAANG interviews twice (The interviews were in cities I wanted to visit, so why not)passed both times and even accepted the offer the second time. In the end I wasn't a very good fit for the company.

What happened if you don't mind me asking? (I might be in a similar situation lol. I just started a FANG on a team of all Indian ppl and I'm the only white dude and I can barely understand the speech much less the code)

I worked with games development most of my life and then went to web development in this change, for some reason I ended up in position a bit more senior than what I was prepared for in this domain. I thought the job would be fine, specially being my second successful interview with the company, I thought their processes meant that I was good enough for the job they offered me.

Is game development as tough regarding WLB as people say? It's always been one of my dreams to work in Game Dev but not sure if I should keep it as a hobby or give it a shot. Are there sane Game Dev co's with decent WLB?

I mostly worked on smaller studios and for the last 6 years of that career I was working in my own company, I only had WLB problems in my first 2 years, but I never worked in AAA

Do any of the interview questions you had, or studied for, have anything to do with your day to day job?

I got a job as a front end developer. We had a ton of streaming data and needed to index it in the front end.

The "right" solution would be to fix it in the back end so that we didn't fetch all that data when it wasn't needed, but that wasn't possible because of the horrific project/product management.

So I had to build binary search trees to index the data so we could work with it fast enough to have a reasonable user experience.

So yeah -- you will need some of this stuff. You're constantly going to be searching for things, reversing things, looking for patterns.

it won't be so clear and abstract, as a leetcode puzzle but the reason why I had to write that binary tree was because whoever came before had clearly never considered using one and was doing everything completely wrong. It was a disaster and made the codebase insane.

If they filtered in the hiring process for people who know the basics then things would have been a lot more performant, and they wouldn't have burned so much time working around the performance issues caused by his terrible solution.

Interesting, thanks for that.

I would add to this that while knowledge of computer science algorithms and data structures is _absolutely_ useful, being able to implement a binary search tree RIGHT NOW, in under an hour, is not necessary for... pretty much any job anywhere.

I absolutely agree that hiring folks who are aware of the basics is important -- essentially, you want engineers who are aware of the space of possibilities.

In my view, what makes folks dislike typical coding interviews is that in the real world, what you need is a solid understanding of what algorithms exist and when to use them/what to look for, rather than the knowledge of how to build one on-the-fly.

To solve the issue you described, you don't need to know offhand how to implement a binary search tree on a whiteboard. You do need to know how to identify indexing as a bottleneck, and how to broadly think about a solution. You could then search for indexing strategies and, having studied them at some point in the past, you'd be able to pretty quickly refresh your memory and find the one that's a good fit for the problem at hand.

For this reason, I've always thought these exercises would be much better off as essentially "open book" rather than real-time whiteboarding problems -- because that reflects how engineers actually work. That's also what I've pushed for in my own workplaces, and we've had good success finding talented folks, and heard positive feedback about this aspect of the process.

What kind of practice did you do? Leetcode exercises or something else?

I've worked at 75% of FANGs. My method was basically just bone up on algorithms and data structures by going onto the Wikipedia list and trying to implement them on pen and paper. Practice thinking out loud, practice space management(no backspacing on pen and paper). Be honest if you've heard a question before. Know how to analyze runtime of your algorithms.

I chose to interview in python, even though I know other languages better, because it is fairly dense relative to, say java or c++ and easier to write by hand.

> Be honest if you've heard a question before

Why so?

If you waste your only evaluation period on a question obviously regurgitating an answer, interviewers notice and can't evaluate you fairly.

If they just want to watch you program, they might still ask you to go ahead, at least you've put the ball in their court.

If they think you weren't forthcoming, that might be an automatic rejection

> can't evaluate you fairly.

What's the meaning of fair? Several people admit you need to prepare, and then it seems you should not prepare too much? It looks like a date.

I can't reply to the sibling commenters but providing a contrarian opinion. I'm an interviewer and there's no way for me to tell who is just really good vs who has seen the exact question before. Telling the interviewer just gives them information that goes against you so I wouldn't recommend doing this.

I guess, it is okay telling your interviewer depending on your comfort level, as most never change the question regardless. That said (and as another commentator points out [0]), if an interviewer asks you call out questions you know of from before, then you most certainly should (unless you can sell the bluff...).

[0] https://news.ycombinator.com/item?id=31067269

Does honesty go against you in interviews these days? I'm getting old.

It does in this case yes, how would you see it as an advantage? You have to pass X number of algo/data structure, your being honest will only make them find a different question. If you get the different question wrong guess what you are out. It is what it is.

If the hiring process is set to punish honesty, then maybe by being honest you avoid working at places with such a process and consequently with the people who passed it?

Note that I'm not saying that FAANGs are like that (I have no opinion on that).

It's not punished but it's not massively rewarded either

how would you see it as an advantage?

How would I see presenting myself as a trustworthy employee who won't pull the wool over your eyes? Pretty favourably tbh.

I am not denying honesty won't hurt, and probably would help you. But again, the interviewer is going to have to get a replacement question...you have a reduced chance of solving it correctly.

I have a question that I've been asking for years that is somewhat domain-specific, but should be answerable by anyone who knows what a set/map are. About 20% of people get past "stage 1" within 5 minutes, usually people with finance/trading experience, and some will say "oh, I implemented something just like this a few years ago." Some people take 30 minutes and do ok, and some never finish. When someone tells me they have had this problem before, I tell them it should be really easy for them then, and we can move on to more interesting things!

For those that DO get past stage one, it is a boss with multiple health bars. We talk about what improvements could be made, then what improvements on THOSE improvements. We keep digging until we are out of ideas. The best candidates are the ones that stump me, or introduce new ideas. I present this problem as a pair problem solving challenge, so there is no one right answer, and has a lot of back and forth.

I always prefer to say "I've seen a similar problem before, let's see if I can remember how to solve that" or "let's see if the same kind of approach works here".

This is a balanced approach. It's honest: you probably haven't heard the exact question word for word, even if you can't recognize the differences. And it's actually what they're selecting for in the first place. They're not expecting you to invent the concept of a binary tree, or whatever, but to know it exists and how to implement it and to recognize where it might be the right concept to apply.

I've never had any interviewer say "OK forget it, let me give you another question where you'll never have seen something similar".

There is no way for you to tell if someone is really good (as is) vs who has grinded leetcode since the process (which assumes prior preps) a priori discards the first group.

You assume that all successful applicants "grind leetcode". What makes you so certain?

Leetcode grinding is suggested as necessary by everyone, including recruiters (in-house or 3rd party), hiring managers, prospective colleagues. You can see comments from devs at FANGs even in this thread.

I understand that this is a common belief, but I don't agree that it is strictly true; it is contrary to my own experience, both as an applicant and as an interviewer.

Note that we are talking about a particular interview process which uses CP (Competitive Programming). This is common at FAG (Lets exclude Netflix since as I've heard they don't do it) and at companies that copycat the process. Of course, there are many other places that don't do it.

We are? What is "competitive programming", and where did it come into this thread? All I see upstream from your comment is a discussion of Google interviews and large tech company interviews generally. I don't recall anything particularly "competitive" when I interviewed at Google (or Facebook, or Microsoft, or...); they just had me solve problems, as usual.

That's the point, CP != CS (Computer Science). It is a separate discipline/subject with its own trivia knowledge & tricks. CP uses Computer Science the same way as e.g. Physics or Biology use Math. And CP problems are used in those interviews. https://en.wikipedia.org/wiki/Competitive_programming#Benefi...

I ask again: what makes you so certain? This does not correspond to my experience with tech interviews.

Again, a common knowledge, including the above mentioned Wikipedia article. Good for you if that was not your experience :)

I was thankfully asked (in the interview, not just assuming I’d been asked the Q before).

Question was: write code to determine if the stack grows up or down. I’d been writing computer games for several years and smashed it (after telling the interviewer that it would be necessarily technically rely on undefined behavior) and the interviewer somewhat dismissively said “you should have told me someone else asked you this question” “What are you talking about? This is just an easy question.”

I guess you’re screwed at that point as now they also think you’re a liar. No way out.

I got hired from that loop. Same guy challenged my failure to write ifdef include guards on my header file on the whiteboard. My answer of “oh, you’re right; I’ve configured emacs to automatically insert those, so I don’t have to think about it” seemed to more than satisfy him and we ended up being close as colleagues for several years.

> “you should have told me someone else asked you this question”

At what point do we end up saying "my current employer 'asked' me this question, because it's part of my day to day job..."? At some point you have some experience in certain areas that you just 'know', and it's not some sneaky "oh I crammed leetcode for 3 weeks!" tactic.

Yes, it seems to me that if interviewers are going to be annoyed about this then they should stop using leetcode and/or using generic interview questions.

Red flag. People forget interviews are a two way street.

You ask a Leetcode question, and then assume the candidate is lying without even asking "have you seen this question before?".

A good answer would be "no, but when I was working on <game title> which is a game that has to run on many different consoles including <console>, we had an interesting but where <weird stuff happened> and the stack cookie I inserted to test if it's a buffer overflow remained pristine. After a lot of debugging, at the stage where I started suspecting it's a compiler bug and inspecting raw opcodes, I was looking at my debugger and noticed the stack pointer was lower than the stack cookie address and understood this CPU has a stack that grows the other way than I'm used to".

Instant hire if I was interviewing you.

It is an easy question for someone that understands pointers. I suspect those sorts of folks are becoming few and far between.

The whole leetcode approach to interviewing is basically

1. interviewers asking questions that they wouldn't be able to solve 2. interviewees pretending to solve on the spot things they've memorized 3. interviewers pretending to believe them

When you’re an interviewer and you’ve asked the same question enough (tens of time is plenty, really) you’ve seen it all and it’s really obvious when someone has seen the question before.

People who haven’t seen the question before always stumble somewhere. There’s something they didn’t notice at first and need to account for, their solution is not structured optimally for the follow ups, they iterate over some possible solutions while thinking out loud to eliminate them etc.

It’s honestly not that hard to tell when someone is pretending they haven’t seen it before.

As an interviewer I find it easy to catch candidates who are regurgitating a memorized answer, but catching people who know the answer in-and-out is really hard. I've had the exact same experiences on the interviewing side of the table as well.

I think interviewers tend to overestimate their ability to catch people who have seen the question before and miss on tons of candidates who are good at answering seen questions.

It's actually not all that hard to stumble on the question at predictable spots. My process for solving a question I've seen before and one that is new to me actually doesn't differ all that much: as several clarifying questions at the start to confirm I understand the problem, then write a quick-n'-dirty solution as fast as possible. In this first pass I usually don't pay too much attention to things like bounds, which you wouldn't memorize anyways even if you knew the solution. Then I run a pass to polish the solution up, present it, then think of edge cases and how I'd test the solution. The only real difference is that I very rarely pretend not to have seen a question (I have done it twice, both in situations where it was clear the interviewer was out of questions and running down a list of things from memory, and I just wanted to humor them). You'd think that I'd be more sure of myself when I actually know the answer, but when I don't actually know what I'm doing I will usually come to the answer as I ask those clarifying questions and start stating the facts that usually score "this person at least has the gist of the problem" points, like "oh this is a graph and we're doing something with distance, let's see if Dijkstra is the right thing to apply".

I don’t work at a FAANG but I hope to one day. But I do interview for Data Engineering roles from time to time for my and other teams. I think by saying you’ve seen this problem before it shows honesty. It shows character. Those are points that, at least for me, are positives and I like to hear it. I’ll still go ahead with the question just to make sure but if something typically takes 30 mins and you finish in 10 then I’ll move on to a different question to fill the gap.

Interviewer here.

When a programmer goes through a question fast and jumps right into the solution, there is two possibilities. One is candidate already knows the question. Two is candidate is exceptional programmer. If the other part of the interview doesn't match up to this, then we will assume the first case.

I definitely place candidates who are being honest upfront above others. They will be reliable and trustworthy, which is important quality for programmers.

First principle that every interview guide teaches you is to take time to understand the questions. A candidate who has spent months preparing for a FAANG interview will highly unlikely rush into answering a question, even if they have seen the question before.

One reason for folks to jump right into the solution is the interview anxiety - which is both due to lack of practice and trying to compare oneself to folks preparing for months ahead of interviews.

As someone who has done a lot of interviewing from the interviewer side at my current FAANG, it's usually obvious when someone has seen the question before - in the debriefs, the candidates who were honest have it noted out loud to the debrief panel & it makes them look good due to integrity, and those who were dishonest also get it noted as a significant negative.

Perhaps I evaluate differently than a lot of interviewers, but I'm primarily interested in figuring out if a candidate has the traits/skills that I'm looking for in a potential coworker - for us, the traits matter more than the skills even.

It's better to fake having not seen it than failing an unseen before question. No amount of honesty points will save you from failing a question. Sure if you are good enough that you pass unseen questions regularly go with honesty. If you are not, better to fake it. If FAANGs have problems with candidates knowing the questions maybe they should think of different interviews?

My coding question I give (in the occasion I am called upon to ask one to candidates) typically is not particularly unique, but practical - if someone wants to study for it, it doesn't really give them a notable edge given all the possible branching questions. The most notable tell if people did see the question prior though is if they jump right into talking about the problem without asking qualifying questions, implying that they are quite familiar with the problem & don't need me to qualify it & don't think to verify scenarios with me.

For my non-coding problems, I just create it from scratch depending on the position/needs & spend a bit of time navigating the scenario myself and store the question in my notes.

As to failing a question, failing a single question isn't necessarily a deal breaker in itself - it's showing a pattern of not meeting the bar that is. I may rate someone a 2 out of 4 if they didn't go into sufficient depth in a particular question I asked, but I probably won't stay in the way of hiring them if they did ok otherwise and that failure was just an aberration. Loss of integrity is perception that is likely to sour people on any upside of hiring though, and overcoming that bar is incredibly difficult - if someone is clearly rehearsed on a particular question and is dishonest about it, they're probably not getting a 3 or 4.

I've failed enough interviews to know that failing a question is almost always a failing interview. There are enough candidates that will get the question right and one of them will be hired, not me. Honesty aside.To be clear I am talking about FAANGs or famous startups that get hundreds of candidates per position.

I'm a senior engineer at a FAANG who is close to reaching staff by promotion - I've conducted enough interviews over the past 5 years at my current company to at least understand how my org operates.

How the process for us works is if you are a strong enough candidate, we have no qualms giving multiple offers simultaneously and working out the reqs afterwards, even borrowing against a future one if we have to. How we evaluate is probably also a lot different & more thoughtful than a lot of candidates realize - we're discussing leadership traits, strengths & weaknesses, and skills in our debriefs and what we all observed about the candidate in our sessions. No candidate does perfect in any given session - even candidates who I have given 4s for have slipped up or had negatives observed.

I think you are an outlier then. What leadership traits do you discover when forcing candidates to do BFS on a binary tree or similar questions? Why would you take someone who said he knows question A and then moved on to bomb question B when you have 30 candidates who solves question A? Ok no one does perfect but bombing a question seriously hinders your prospects. If you see a question you know or semi know you gotta be very silly to say it. People don't drill Leetcode to say oh sorry I saw this one already. Nope. In fact even remembering hundreds of questions and being able to solve them under stress is hard enough. Many FAANG candidates are just brilliant and don't really need much preparation to get accepted. But others are normal smart but spent months preparing by going through algorithm questions. It's quite certain they come up with 1-2 question they have seen before and this enables them to pass. If it wasn't the case no one would subscribe to Leetcode....

My org is generally anti-leetcode so I can’t speak for your experiences - the only data structure/algorithm questions you may encounter in an interview with my org is likely practical questions (i.e. problems we have had to solve on the job).

I’m usually not even asking a coding question in my session - I set up a practical/common problem beforehand and we explore the scenario together. I can assure you that many candidates don’t pass my session necessarily, even if they have proven in other sessions to be brilliant coders - I’m not looking for technical brilliance in most of the interviews I give, and neither are the hiring managers I work with. To me, focusing on the coding is most important on the technical screen, not the full panel - once you reach the full panel, your goal is to demonstrate technical leadership, which includes expertise in knowledge, coding competency, focus on UX, and some other areas for more senior roles (conflict management, responsibility, navigating different stakeholders for product/project decisions, etc.).

If all your focus is in is just the coding questions, you’ve likely already set yourself up for failure.

I don't think I've ever been asked a question at a FAANG interview that isn't in one of the (legion of) "FAANG interview prep" books/sites that ex-FAANG engineers and interviewees constantly publish and update.

Your criteria here is likely unfairly dinging pretty much everyone who has done significant interview prep.

But significant interview prep hurts the primary job of the interviewer, which is to evaluate whether candidates would work well within the team/org - I don't mind if candidates do preparation or not, but the point isn't to try to find people who game the process, the point is to find out if a candidate is likely to be successful in the role. The problem is oftentimes it turns out that a lot of candidates who focus too heavily on preparation fail to demonstrate the qualities needed to be successful.

I don't give a particularly hard interview, candidates and interviewers who pair with me are all pretty happy with the session & almost always leave relaxed, which is usually targeted towards a certain set of leadership traits (and occasionally I'll do coding screens as well although I've been put on those less in general). My interview feedback also has generally been corroborated by other interviewers in debriefs as well, so it's not like it's in crazy land.

FAANG recruiters literally send the candidates emails that say things like "you should study Cracking the Code Interview beforehand". No, they aren't supposed to, and no, the incentives aren't set up correctly to prevent it (I've received these emails from Google, FB, and Amazon recruiters in the past).

We can't really go around penalizing behaviours that the recruiting side of the house is actively encouraging shrug

I can definitely tell you that our recruiters do not tell candidates to do that.

If this is the first time, at the same company, it (probably) does not matter, that much.

When I was an interviewer for a (second) technical phone interview at Google, one candidate performed pretty bad at one question. Later that week, or maybe the following week, that very interview was in the pre-HC meeting and one of the other pHC members pointed out that I had repeated a question from the first phone interview. At which point I pointed out that I had not been provided a list of previously asked questions, the candidate had not highlighted it, and the candidate's response was still sub-par.

Not mentioning the repeat counts against character and responsibility. Not performing better at a repeat question a week later counts against competence.

That was an easy "let's not bring this candidate on-site" decision.

I did problems from Elements of Programming Interviews, on paper, while talking out loud about what I was doing.

I would have used a white board if I had one, or a google doc (non-ide text editor) if I were practicing for a remote interview loop.

I tried to do one problem a day, capped at 45 minutes, plus a bit of time to confirm my answer if I got it, or understand the answer of I did not.

My goal was to practice things the interviewer is looking for, in a setting that is as close as possible to an interview (no ide, time pressure, etc.)

* Alternative approaches to solve the problem.

* Test cases, including walking through some.

* Runtime/space complexity analysis.

I didn’t have a good study plan for design interviews, but I’m better at YOLOing those :)

Reply on news.ycombinator.com