|
Good Will Hrunting posted:I get it, not everyone can work at a FAANG and this process is "probably fine" for those companies, but the level of ubiquity we've reached in companies copying them is annoying. The trouble being is that *it means nothing* - except that someone can memorize these specific toy problems. The sooner we, as a professional community, stop blowing the FAANGs as some sort of 'the best of the best' - and therefore assuming their hiring practices lead to having great developers. Anecdotal experience largely leads me to believe that there's just a distribution in play and that distribution is absolutely, completely blind, to human concepts like "place of employment". Not sure how I'd describe that distribution, today I'm cynical, so I'd say that for any random sampling of developers - 20% do 80% of the work, and 20% are responsible for 80% of bugs and maintainability issues, and there's some overlap in those categories.
|
#
?
Apr 14, 2021 11:34
|
|
|
|
| # ? May 24, 2024 13:44 |
|
|
Cuntpunch posted:The trouble being is that *it means nothing* - except that someone can memorize these specific I don't want to wander too deep down the class divide discussion road, but getting into/graduating from medical school follows a similar pattern that just selects for people who 1. Has the temperament/raw ability to memorize tons of trivia that's not specific to their specialization* 2. Selects for people whose families can support them for months if not years to do nothing but sit and memorize things If you're working 2 jobs and taking care of young kids you'll never have the time/focus to do #2 even if you're particularly good at #1 I'm sure there's some rough correlation between memorization and analytical ability**, but at the same time I know/work with a lot of developers from "non traditional backgrounds" who are really talented but would fail a traditional whiteboard interview * What proctologist needs to know the seven bones that ring your eye? ** I guess? Seems like bullshit but this system wouldn't have survived for 100+ years if it didn't work
|
|
#
?
Apr 14, 2021 15:45
|
|
|
Cuntpunch posted:The trouble being is that *it means nothing* - except that someone can memorize these specific toy problems. The sooner we, as a professional community, stop blowing the FAANGs as some sort of 'the best of the best' - and therefore assuming their hiring practices lead to having great developers. if you have a problem effectively doin satisfiability youre gonna have an 80 20 sorta power law pop up cuz satisfiability is itself a spin glass w second order phase transition and dynamic and static 1 step replica symmetry breaking. so its just a universal (and universality is a technical term in physics) fact of systems that try to satisfy some requirements one of the grand interfaces betw physics and cs. they got simulated annealing out of it in the next paper after that. real ironically, they dont teach you this at any place but school
|
|
#
?
Apr 14, 2021 16:25
|
|
|
Cuntpunch posted:The trouble being is that *it means nothing* I think it means something. It weeds out people like myself that are too dumb, even after a lot of practice, to solve problems like this quickly on a whiteboard. People have to accept that sometimes... they might just not have the brainpower to do these problems. It certainly says something about that and the ability to quickly code up solutions to algorithms, and effectively adjust them, adapting under immense pressure.
|
|
#
?
Apr 14, 2021 16:36
|
|
|
The FAANG interview process definitely measures something, but it doesn't measure it all that well, and what it measures isn't all that important for working as a developer.
|
|
#
?
Apr 14, 2021 16:38
|
|
|
Debating the effectiveness of the interview process is all well and good, but the guy's question was how to prepare for an interview. It doesn't matter if the interview process is good or bad, it's the current reality. Whether you like the process or not isn't going to help this guy get a job
|
|
#
?
Apr 14, 2021 16:48
|
|
|
ultrafilter posted:The FAANG interview process definitely measures something, but it doesn't measure it all that well, and what it measures isn't all that important for working as a developer. dan luu notes that he finds algorithmic poo poo all the time that all his algo solvin coworkers missed lol im guessin its cuz he actually learned this poo poo and likes it instead of learnin it to jump thru hoops the fundamental problem is lack of trust and bad faith. both are very well warranted in both sides. just practice more, endure the suck i guess bob dobbs is dead fucked around with this message at 16:52 on Apr 14, 2021 |
|
#
?
Apr 14, 2021 16:49
|
|
|
Good Will Hrunting posted:I think it means something. It weeds out people like myself that are too dumb, even after a lot of practice, to solve problems like this quickly on a whiteboard. People have to accept that sometimes... they might just not have the brainpower to do these problems. It certainly says something about that and the ability to quickly code up solutions to algorithms, and effectively adjust them, adapting under immense pressure. It's not "brainpower," it's stress. You aren't dumb, you just have a completely normal stress reaction that the FAANG interview selects against. quote:Through a happy accident, the software industry has seemingly reinvented a crude yet effective instrument for reliably introducing stress in subjects, which typically manifests as performance anxiety [75]. A technical interview has an uncanny resemblance to the trier social stress test [39], a procedure used for decades by psychologists and is the best known “gold standard” procedure [1] for the sole purpose of reliably inducing stress. The trier social stress test involves having a subject prepare and then deliver an interview-style presentation and perform mental arithmetic, all in front of an audience. Alone, none of these actions consistently induce stress in a subject; however, the unique combination of cognitive-demanding tasks with a social-evaluative threat (essentially being watched) is consistent and powerful. If a technical interview is essentially a de facto version of a trier social stress test, then the implications can be profound. Rather than measuring the few that answer correctly in a timely manner, companies are most likely measuring the ability of the few who perform well under stress [30].
|
|
#
?
Apr 14, 2021 17:00
|
|
|
stress in the trier stress test can be trained to be lower. go train for the stress. practice like the test or harder. unfavorable peeps lookin over you, tennis balls thrown at you while you're doin stuff, etc etc
|
|
#
?
Apr 14, 2021 17:05
|
|
|
downout posted:I've got an tech interview coming up soon with a larger tech company. I've got some anxiety, but I'm pretty excited too. This could be a great career move. I'll bite and give actual faang interview advice. This applies to Google style interviews but not places that do the huge take home interviews, ymmv. The point of them is to see how you think about computation problems and so talking out loud as you think is something you need to do (honestly more than getting the right answer immediately). If that's not natural you really need to practice that. Faang style interviews will have a couple screening interviews for basic coding but the main batch are a lot less about it than the whole path of given an abstract problem come up with a good way to do it and then write what probably won't be much code to actually do it. You don't have to get there without making mistakes especially if you spot them yourself, but if you bounce back well from having them pointed out it's a positive mark. So brush up on algorithms, not memorizing specific ones (though you should be comfortable with basic stuff like searching and sorting) but actually starting from a word problem and solving it to code. When I help friends prep I'll grab questions from the back of algorithms books without telling them the chapter, dynamic programming word problems are great practice imo. Yes they're toy problems but they're just to see how you work through solving a problem computationally, they're not brainteasers. I'm a big algorithms nerd but I always prefer the structure of: do an example to make sure you understand the question, come up with a naive solution, understand the performance of that naive solution and why it sucks, and then come up with something better. It's a bit to fit into an hour but it'll get you there. Listen to hints and prompting questions from the other person, they're usually not jerks.
|
|
#
?
Apr 14, 2021 17:15
|
|
|
I've done and conducted FAANG style interviews, and all criticism aside (all valid, but different topic), the top pieces of advice I can give is to ask clarifying questions, probe the boundaries of the question for edge cases, vocalize your thought process, and write down your notes/diagrams/pseudocode as you go. The number of candidates that I have interviewed that don't tell me what they are thinking, don't ask questions about what they are unclear on, and then also leave no notes of any kind on the whiteboard/codepad is staggering, despite much encouragement even in the moment. These interviews are not 1-way I ask you a question and then stare at you blankly for 45 minutes situation. That's agonizingly terrible for both of us. I want it to be a back and forth exercise, where you tell me what you're thinking and how you're approaching the problem and I can gently course correct you or tell you what is in bounds and what is not in bounds. My questions are designed to have multiple paths to follow depending on which direction that candidate takes it, all equally valid. But you have to talk to me. Lastly, the interview consensus meeting often happens 24-72 hours after the interview. I may have interviewed several candidates this week. Unless you made a lasting, stellar impression in the interview, I'm largely going to rely on my notes and whatever you wrote down. If you don't write anything down you're doing yourself a huge disservice when I go back to look at them the next day to write up my summary. The actual technical side of the questions tends to be much less difficult or scary than all those hard leetcode problems. It's about your problem solving approach and process and how you face something unknown. You need to focus on demonstrating that very clearly. I've been asking the same handful of tech questions of all candidates I've interviewed in the past couple years, the same questions that my peers also ask in their interviews. The questions are even posted in candidate reviews on Glassdoor. We know the questions may be "in the wild", but we don't care. The question itself is largely not the point, it's how you work through it and demonstrate knowledge. I give a stronger thumbs up to someone who may have stumbled a little bit but was asking good questions and being vocal about their approach than to someone who just whipped off a perfect solution in silence. Guinness fucked around with this message at 17:36 on Apr 14, 2021 |
|
#
?
Apr 14, 2021 17:29
|
|
|
The last time I interviewed at Google, it was me talking to a wall. My interviewer kept answering "I don't know" when I asked for clarification on the problem that was copy and pasted to a Google doc and formatted in such a way that the examples were borderline illegible. It was 2d array traversal and the matrix that they gave me as an example input was so hosed up I had to copy it to a piece of paper to understand what they were getting at. Sure the problem itself was "hard" but I was already really flustered trying to understand it before even getting started. I'm sympathetic towards the interviewer because it was clear that they had communication problems (the question was written out really poorly too) but it's frustrating to prep and be hit with something like that. It really soured my experience. The first time I interviewed it was a simple DP problem or greedy algo problem - I don't remember - but actually one of the better interview experiences and my interviewer was exactly like you said above. I had already had an offer somewhere else though so I didn't go through with further steps.
|
|
#
?
Apr 14, 2021 17:34
|
|
|
the reason why peeps dont talk to you is cuz of the stress lol (in reply to guinness) practice the stress if you have the problems relatively down there is no actual consistency in interviews at any company bigger than 10 peeps (maybe like, 50). one side of it all bein a market of lemons bob dobbs is dead fucked around with this message at 17:44 on Apr 14, 2021 |
|
#
?
Apr 14, 2021 17:36
|
|
|
Yeah and I bet most interviewers at Google are at least decent. The problem is you have to pass 5 or 6 or something to get an offer. I have zero interest in anything besides G in the FAANG group so I'm trying to decide how best to prep this time around.
|
|
#
?
Apr 14, 2021 17:48
|
|
|
Good Will Hrunting posted:Yeah and I bet most interviewers at Google are at least decent. The problem is you have to pass 5 or 6 or something to get an offer. There's definitely lovely people interviewing anywhere, and that's the worst for that style of interviews but it's not "X% is passing else no" system fwiw. It depends a lot on what the interviewer wrote and even their own past interview distributions (a yes from someone who is well calibrated means more than a no from someone who isn't)
|
|
#
?
Apr 14, 2021 17:53
|
|
|
bob dobbs is dead posted:dan luu notes that he finds algorithmic poo poo all the time that all his algo solvin coworkers missed lol That may well be part of it, but also keep in mind that once you start actually working at a company, your priorities become completely different. O(n^3) or worse algorithms show up all the time because what matters is getting the product/feature working ASAP, not making it pretty or efficient. Hardware is cheap, developers are expensive. Good Will Hrunting posted:Yeah and I bet most interviewers at Google are at least decent. The problem is you have to pass 5 or 6 or something to get an offer. Yep. Unfortunately, even if you do everything right, there's always a nontrivial chance that your application gets torpedoed for reasons outside your control. lovely/biased interviewers, vagaries of HR, managerial nepotism, etc. can all conspire against you.
|
|
#
?
Apr 14, 2021 18:01
|
|
|
I have a bit of an internal monologue when solving problems, but I've generally learned to not verbalize it unless doing mentoring/pair type work. Based on some of the suggestions upthread I think I'll try to verbalize more during this, it often helps me solve the problems or come to realizations faster anyways. I really like the suggestion to make notes where possible, that may be more difficult in remote interviews, but I'll watch for opportunities. All of these suggestions and discussion have been really helpful. It's given me some direction on where to spend time brushing up. I might try to setup a distraction thing with some people this evening to try and better prepare for it.
|
|
#
?
Apr 14, 2021 18:20
|
|
|
ultrafilter posted:The FAANG interview process definitely measures something, but it doesn't measure it all that well, and what it measures isn't all that important for working as a developer. I agree that it's almost certainly not measuring the correct thing, but the entire process is effective at not hiring the bottom X% in my experience though it does so by excluding a bunch of good people as well. That by the way is the major difference not the whole 10x bullshit. asur fucked around with this message at 18:47 on Apr 14, 2021 |
|
#
?
Apr 14, 2021 18:34
|
|
|
the original claim of the academic peeps who were claiming the 10x stuff was comparing the worst to the best programmers, whereas what the dipshits hiring twisted it to was comparing the median to the best programmers. makes marginally more sense when you're comparing jeff dean to the non-fizzbuzzer
|
|
#
?
Apr 14, 2021 18:38
|
|
|
And in my anecdotal experience some literal garbage tier engineers have gone on to Google after being laid off from my companies. That's the only reason I'm still considering the process again!
|
|
#
?
Apr 14, 2021 18:40
|
|
|
Well. Got laid off last week. They're giving me a month's pay as severance, and I've got about a month and a half's salary in my savings account, plus whatever unemployment gives me, so I'm not super worried right now. My dumb job had me back in the office since Memorial Day last year, and I gotta say, those weeks between mid-March when lockdown first started and Memorial Day was some of the best time of my life. I never want to go back into an office full time ever again. Any advice for looking for remote work? I know Indeed has a Remote filter, but many of the postings still seem to be Temporary. Is the interview process generally any different? This might vary from company to company, but am I expected to use my own desktop PC or might the employer send me a laptop? Should I get a webcam for interviewing's sake or I guess I can just do most of it on my phone probably? Are take-home projects more likely for remote jobs? I hate those things with a burning passion and am sorely tempted to essentially tell them, "I've been a professional developer for 9 years. You can ask me any question, but I'm not wasting my personal time just to demonstrate that my resume isn't bullshit" in nicer terms 
|
|
#
?
Apr 14, 2021 20:29
|
|
|
Sab669 posted:Well. Got laid off last week. They're giving me a month's pay as severance, and I've got about a month and a half's salary in my savings account, plus whatever unemployment gives me, so I'm not super worried right now. I don't know about any of the rest of it but this twitter thread has several posts about where to find remote jobs: https://twitter.com/RandallKanna/status/1285322775360532480?s=20
|
|
#
?
Apr 14, 2021 20:32
|
|
|
Sab669 posted:Well. Got laid off last week. They're giving me a month's pay as severance, and I've got about a month and a half's salary in my savings account, plus whatever unemployment gives me, so I'm not super worried right now. If you're not a contractor, they should send you hardware. And definitely get a webcam for interviewing.
|
|
#
?
Apr 14, 2021 20:32
|
|
|
Jose Valasquez posted:I don't know about any of the rest of it but this twitter thread has several posts about where to find remote jobs: That is a long list, hot drat. Thanks raminasi posted:If you're not a contractor, they should send you hardware. And definitely get a webcam for interviewing. I was thinking I'd look for W2 work, but maybe I should consider self employment. I guess it varies from Statte to State, but any idea what to expect to pay in taxes going that route?
|
|
#
?
Apr 14, 2021 21:51
|
|
|
Sab669 posted:Well. Got laid off last week. They're giving me a month's pay as severance, and I've got about a month and a half's salary in my savings account, plus whatever unemployment gives me, so I'm not super worried right now. I'd recommend finding companies using the stack you are familiar with. Then go to the companies' career sites. Many are including 'remote' in the description/title (that are hopefully searchable) or have a 'remote' or 'virtual' location once you select the country. It's a bit trial and error to find them sometimes. Indeed and the other sites seem to have put in a 'Remote' location now as well. And often the descriptions or title will state remote or not. You'll need a webcam and probably some dev tools installed for the remote interviews.
|
|
#
?
Apr 14, 2021 21:56
|
|
|
Sab669 posted:Are take-home projects more likely for remote jobs? I hate those things with a burning passion and am sorely tempted to essentially tell them, "I've been a professional developer for 9 years. You can ask me any question, but I'm not wasting my personal time just to demonstrate that my resume isn't bullshit" in nicer terms I hate these too, yet I’m not in your position. I’m posting for conversation on the topic, not because I have good advice. Even if you do the homework there’s no guarantee they’ll even review it. I’ve done homework only to be told “they’ve instated a hiring freeze.” Even in the case an unemployed candidate, doing homework just for the unenforceable promise of further consideration costs time that can be spent on other applications. Whiteboarding isn’t perfect but at least it requires an investment from both sides. It would cost money for companies to offer in bad faith. dividertabs fucked around with this message at 04:11 on Apr 15, 2021 |
|
#
?
Apr 15, 2021 04:08
|
|
|
humorously, companies can and will spend 1-3 grand a pop on all day interviews in bad faith yes
|
|
#
?
Apr 15, 2021 04:10
|
|
|
bob dobbs is dead posted:humorously, companies can and will spend 1-3 grand a pop on all day interviews in bad faith yes As long as they’re burning money that way it should feel better. Like those people who keep phone scammers on the line as long as possible.
|
|
#
?
Apr 15, 2021 04:28
|
|
|
Sab669 posted:I was thinking I'd look for W2 work, but maybe I should consider self employment. I guess it varies from Statte to State, but any idea what to expect to pay in taxes going that route? The big tax difference between W2 vs self-employment is in social security and medicare taxes. The IRS wants 15.3% of your gross for SS/medicare. If you're W2, this is split between you and the employer; you personally only pay 7.65% and your employer pays the other half. If you're 1099 you have to pony up the entire 15.3%. Also remember that as 1099 you won't get vacation time or other benefits so make sure you have a rate that allows you the opportunity to take some (unpaid) vacation time and also to pay for health insurance.
|
|
#
?
Apr 15, 2021 04:29
|
|
|
dividertabs posted:I hate these too, yet I’m not in your position. I’m posting for conversation on the topic, not because I have good advice. My current place has homework, but it pays hourly and has a lot of back and forth with a dev and a product manager so it absolutely gets looked at. Employers can invest in homework too. (I realize most places don't.) edit: and to contribute to the op's question, this place is fully remote.
|
|
#
?
Apr 15, 2021 04:54
|
|
|
pokeyman posted:My current place has homework, but it pays hourly pokeyman posted:Employers can invest in homework too. (I realize most places don't.) In reality I think most places invest something in these homework assignments. There's not that much incentive to waste candidates' time. But candidates rarely know until after they've invested some time (your place, paying for it, being one of the exceptions).
|
|
#
?
Apr 15, 2021 05:30
|
|
|
I'd rather spend time on a homework project than memorizing leetcode problems, any day.
Ralith fucked around with this message at 17:02 on Apr 15, 2021 |
|
#
?
Apr 15, 2021 16:49
|
|
|
Normal brain: memorizing leetcode problems Galaxy brain: memoizing leetcode problems
|
|
#
?
Apr 15, 2021 16:57
|
|
|
Ralith posted:I'd rather spend time on a homework project than memorizing leetcode problems, any day. The time spent practicing leetcode carries over to any interview that does algorithm questions. Homework projects are just sunk time that is only applicable to 1 interview
|
|
#
?
Apr 15, 2021 17:06
|
|
|
My takehomes have been timed LeetCode strong Medium or Hard problems where I get a 2-3 hour window to solve them. Amazon did that, Disney, couple startups and a major bank last time I did the process.
|
|
#
?
Apr 15, 2021 17:55
|
|
|
not me. spend a few hours brushing up and i'm good for whatever (within the bounds of luck). homework is a couple hours _per interview_ and eff that it really just depends on preference. i usually do well on system design stuff, so i'd much rather whiteboard that in a 45 minute timebox. the algo questions i got for SWE interviews always tended to be on the order of "use a hash" or "create and traverse a graph". occasionally "put poo poo in a cache and call it dYnMiC pRoGrAmMiNg". now that my title has "security" in it interviews are way less algo-heavy, worst thing i've had since the title change is drawing out how like CBC works or whatever. in general, my thoughts on algo interviews are that you should know the basics with respect to the common data structures and when to apply them to problems. you should be able to pseudocode implementations of them for easy versions. even lovely algo-heavy interviews will often let you iterate through the list and call AddToHeap(el) or whatever. from there you're gonna be fine for a good grip of interviews. you'll also fail some "one dumb trick" interviews, but those are bad and are becoming less common IME. but like graph traversal and stacks and queues and heaps and some flavors of trees and how to use them are just kinda, like, knowing how to use tools. it's true that a lot of webdev stuff doesn't need every one of those tools, but even in that realm my experience is that somebody who knows the difference between a set and a list is way less likely to do dumbass n+1 type stuff when they aren't paying attention. and even in web-space, if you hear "implement autocomplete" and don't think "trie" that's kinda bad. if you can't remember how to implement a trie but can use an imaginary trie library in your interview, that'll get you most or all of the way to success because you know the part that really matters. there's nothing wrong with knowing things that you don't use every day. i know a lot of words that i don't often use, but knowing them lets me choose the correct words when i need to be precise or make a point or whatever. it's good for CS students to take humanities classes. it's good for webdevs to know what a priority queue is. if a SWE doesn't have any understanding of the stuff i talk about above, then they should learn it. and if an interviewer wants something more than that, we're in "one weird trick" territory, and that's also bad (much worse than the former imo)
|
|
#
?
Apr 15, 2021 22:11
|
|
|
Crap I'm sorry, I'll have a spicy chicken combo, Biggie sized, with a Frosty. Yeah, chocolate
|
|
#
?
Apr 15, 2021 22:12
|
|
|
One time on a phone tech screen I got asked to implement a red-black tree and I said "No. I know that they are a way to implement a efficient dictionaries but I cannot implement one for you off the top of my head. If I had to do that on the job and my standard library didn't have one I'd look up how to do it." I was expecting to get bounced for that but they brought me onsite anyway (where I actually got bounced).
|
|
#
?
Apr 15, 2021 22:39
|
|
|
My favorite interview story was the time I got asked to implement an unusual data structure (can't remember exactly what it was, some kind of modified map I think?) and the person who was conducting the interview had to stop and Google the answer to see if I was right.
|
|
#
?
Apr 15, 2021 22:43
|
|
|
|
| # ? May 24, 2024 13:44 |
|
|
It's very hard to define "one weird trick" territory but getting an accepted submission for a pretty good number of LeetCode Medium/Hard questions generally requires something on the verge of "one weird trick" or at least an "aha" type moment where you realize you have to use a second data structure, generally hash table to preserve some sort of mapping, that makes your algo at least slightly (potentially way more) complicated and might introduce nuance to handling earlier cases or program flow control you hadn't considered. Then you might have to scrap your original idea when you realize it wasn't sufficient.
|
|
#
?
Apr 15, 2021 23:11
|
|