My Embarrassing Google Interview

Hey, everybody. Welcome back to Stevey's Tech Talk. Today, I've got some story time for you. I want to tell the story of my Google interview. I think it would be useful if we all told each other our interview stories a little bit more often.

I know that interviewing is so stressful. You spend a lot of time interviewing people and you forget how stressful it is to be on the other side of it. Your brain turns off when you get in front of a whiteboard. You regress on a bunch of things that you thought you knew about, if there's any pressure at all, which there always is. Even if you don't care if you're going to get the job — they're still evaluating you in your profession. So it's like a test to see if you passed at life. And it can get really intense.

And my Google interview was no exception. My Amazon interview was a joke. I interviewed at Amazon in late 1998. Amazon was very, very small at the time — maybe 200, 250 people, in one little building in downtown Seattle.

And my interview was — I'm not going to say they were looking for someone who could fog a mirror, because they had bar raisers and everything and they did hire great people. But for some reason, they decided you were great before they interviewed you. So when I went in, they weren't interviewing me at all. They were just selling. One VP got it in his head that he wanted me on his team, even though I was interviewing for another org. So he spent the whole hour pitching me on his team and how great it was. And they were not happy with him about that, because I guess that was a no-no.

At Google, they do things a little bit different. They don't tell you where you're going to go. You just go into a big pool of candidates who get interviewed and have to pass the Google bar. And then if you make it in, you get a choice of a couple of teams to go to — teams that happen to have open headcount.

When I joined Google in 2005, the choice I got was, they were like, let us know what you want to do. Infinite choice. I could go work on any team in the whole office. And it turned out to be a shockingly difficult decision to choose which team to go to. It was so hard. I'm sure it was hard on everybody else, because they eventually changed it to where they only give you two options. They were getting feedback from people that it was overwhelming.

So anyway, the whole choice thing was crazy, but I wanted to tell you about how I got into Google. And then if we have some time, maybe I'll tell you about what my first week was like. It's one of my best stories. I don't think I've ever properly told the story.

Now, one problem with telling the story is that just to tell it right, I would have to give away too much information about who was involved. And I don't want to dox anybody here. So I'm going to be changing some of the details of this story to provide a nice Gaussian blur over all of the details, so that you can peer into it and try to guess who's involved. But hopefully you won't be able to. It was all in fun.

So, first of all, I applied to Google five times, and it was on the fifth application that I finally got a recruiter to reach out to me. This was in 2004 and 2005. They IPO'd in 2004, so it was just after the IPO. And I was like, oh man, I got to get in on this. I should have gone pre-IPO. I was at Amazon, and all the interns were going to Google and everyone was buzzing about Google. But I waited until after the IPO. And it wasn't until a good year afterwards that they finally noticed me, because everybody wanted to work for Google. They got a million resumes a year. It's impossible to even go through that much data.

So every two to three months I would apply to a different group with a different resume. The first one — radio silence. So the second one, I wrote a completely different resume and changed all the order around of everything I had done and tried to highlight different things and apply for a different team. And I did that five times before they finally noticed me.

And before I go on, I will also note that I have friends at Google who had to interview and then they failed, and then they re-interviewed up to five times. Five is the record that I'm aware of. Because interviewing is not accurate. It's great if you're okay with heavy false negatives — you'll turn anybody away just in case, so only a few people slip through and you keep your bar high, but you turn away a lot of good, qualified people. I know people who could have made a big difference at Google, and instead they wound up advancing Amazon because they didn't get in on their first try.

So anyway, it's onto the interview. I get scheduled for an interview at Google. Finally, the recruiters are, oh my God, we've got to get you in. What an amazing resume. And I'm like, you could have looked at my resume I sent a year ago, or the one three months after that — they were all pretty much the same. So they set me up with some senior interviewers. I had three interviews. It was a small loop, like a half a loop. And they wanted to do that because they weren't sure if I was going to merit having three more interviews.

It's a pretty common tactic for companies who are on the fence. They look at the resume and say, well, they seem pretty good, but let's give them half a loop. And then if they do well, we'll bring them back for the other half. If you go in and you completely tank, well, at least you only had to talk to three people. So they minimize the pain.

The first person was a good interview. The guy is still a good buddy. He was a famous Microsoft person, and asked me a bunch of operating-systems-like questions, a bunch of platform-like questions. Because he knew about my game, and he wanted to ask me about my game. I had mentioned it on my resume. Of course I had seven years of Amazon experience, but no, they wanted to know about my game.

And it was remarkable how penetrating his questions were. He got straight to the heart of my game engine. How did I do scheduling and fair scheduling, and how did I handle resource loading, and what was the player experience while things were waiting? He was walking me through my data structures and my lock synchronization. It was like he had just finished looking at my source code — half a million lines of source code. And he was asking nonchalant questions about this. And I'm like, who are you? How do you know all this stuff?

And he worked on Internet Explorer — I don't want to say in what capacity, but imagine an important one. And of course, the browser is a platform as well, and has all the same scheduling, resource management and multi-tenancy issues. So that one went well.

And my second interview, he asked me more traditional interview questions. He did ask me one that got me really good. The only one that I remember was, he asked me shuffle. Shuffle is, how do you shuffle a deck of cards? And in computing there are a lot of different ways to do it that are wrong — they're not a perfect shuffle, in the sense that what you want is a completely random chance for each card to wind up in each spot.

But when he asked me shuffle, he didn't ask me shuffle. He tricked me. He's like, well, let's say that we've got a QA department. So I'm immediately thinking about tasks. He's primed me for the wrong problem, on purpose. And he's like, we've got all these tests we want to run, but you know how tests have dependencies sometimes, accidentally. You'll run test A, and then you run test B and B fails, but if you don't run A, B runs just fine. You guys have all run into this, where your tests have inter-test dependencies and they're not supposed to. So what we're going to do is run the tests in a random order each time, to catch some of these dependencies naturally during the daily test runs.

So how would you take a set of tests and run them in a random order? What he was really asking was shuffle. And I didn't know — I'm like, oh, random order. And I came up with a bunch of stupid answers. Had it clicked that it was shuffle, then maybe I might've given a little bit better answer.

But I felt bad about it. And I've been telling people ever since then that the number one thing you have to get good at in order to be good at interviewing at a big tech company is problem identification. Being able to know what question they're asking you is half the battle. Because then you can look at it and go, oh, okay, this is a probability theory problem, or this is combinatorics, or this is a computational geometry problem, or this is a machine learning problem. And now you've preloaded your brain with all the right stuff.

So it turned out he didn't care what my answer to shuffle was. He was a good interviewer. The reason he asked me was, he wanted to see how I thought about it. He wanted to see if I would engage with him in asking questions, trying things. So I don't think he cared that I got the answer right.

What he cared about was that I knew why my answers were wrong — which I was more than able to tell him. Yeah, I do this, but of course that gives this card a much higher probability than the others; it's hard to make them all equal. Or I'd come up with a good solution that was very random, but it was too memory intensive or too CPU intensive. And I heard, after the fact, that he liked the way that I conducted the interview as an interviewee. He liked the engagement.

And then there was the real interview. Because the others were just placeholders — there were a bunch of bunny rabbits, and then a Tyrannosaurus Rex. There was a guy at Google, and I'm going to call him Tom, to throw you off the scent. And Tom, in addition to being one of the nicest people you'll ever meet in your entire life, is also a terrifying individual. You are instantly emasculated if you are a human male and walk into a room near him. And it only goes downhill from there. Imagine somebody who's tall, with movie-star good looks, and is so highly decorated in their field that there are no more awards they can get.

Google has this reputation for bringing in the smartest people in the world. Certainly they did for their first 15 or 20 years. The smartest people in the world made a pilgrimage to Google. Google attracted them to be the best, and they did it by making it a better place to work and live than anywhere else. Let's say you're a tenured professor at MIT — that used to be as high as you go in your field. And Google was like, well, it's better here, because you can come work here and there won't be any stress, and you'll be able to work on your field and publish papers, and also we'll pay you a whole bunch of money and have free food and perks. So people came.

And who knows where my third interviewer came from before Google. He was the kind of guy that intimidated everybody by how brilliant he was. Imagine that there are 10 people at Google who are Google, who make the difference. This guy's one of them.

I didn't know that. I had this third interviewer who was a little bit more quiet than the rest. And he asked me a bunch of relatively innocuous questions. Let's say you need to display an ad to a user and you've got a keyword. I think the main problem he asked me was keyword search ad targeting. Given a search query, a bundle of words, and given some number of millions of ad placements that are tied to those words, how would you build a structure that serves the ad quickly for that bundle of words? All the basic stuff I got right — make sure to sort the query so it becomes a single string, and a couple of other things.

But the problem was that, like all good interview questions, his had a lot of depth. One of the things Google likes to ask is questions where your answer should begin with: how important is it that we be accurate? I've seen a lot of Googlers get excited when, sitting at lunch, they ask me a question and my first question back is, how accurate do you want to be? And they'll go, yes, that's right, this candidate wouldn't be right. Well, I was that candidate, on my first Google interview day, on my final interview of the day, with Tom, the crazy, intimidating guy.

He's like a wizard. He's like Dumbledore and we're all like third-year students at Hogwarts. That's a pretty apt comparison. He's Dumbledore, except he looks like Thor. So, very intimidating. And he was asking me these questions. And of course I didn't ask the key important question, which was, how important is it that we be accurate? Because if you ask that, then they get a chance to say, well, think about it — you're serving an ad, what's the worst that can happen? You serve the wrong ad once in a while. So it's okay if it's not a hundred percent accurate. And that opens it up to a whole bunch of new potential solutions that involve probability, or approximations, or heuristics, bloom filters and whatever. You can bring in all kinds of solutions that are mostly accurate.

Well, of course I didn't do any of that. I fell apart. I turned into a blob of jello. Because what I could tell was happening was, he was dumbing the question down. He'd asked me a question and I would know what he's asking, and I'd know approximately what page of the textbook the answer is on, because I kind of sat through it in class, but I wouldn't actually know the answer.

And he would do this thing. You'd say, well, is it combinatorial? Is it factorial? when the answer is combinatorial, and he'd go. For those of you who are listening on the podcast version of this, what I did was my best attempt at an infinitely sad, old, wise deity who's been asked by a small child — the child is asking, do you think I could fly up to the top of that building and then rip the top off of it? And the child is serious. And the deity is looking down at the child with a mixture — let me try to tell you the recipe. The recipe is definitely two parts pity, one part revulsion, two parts amusement, and then nine parts steely resolve.

It's exactly how you would react if a little child, like five years old, walked up to you and said, hey mister, do you think I could lift that car up? And you kind of shook your head at them. That's the look he would give you.

He dumbed the question down, not by double, but by an order of magnitude. And this is what killed me. He would ask me a question that was 10 times easier, and I still wouldn't know the answer to it. And then I'd know I was in trouble, because he'd asked me the 10x dumber version of it. I'd been on the other side of the interview table many times in my career. By this point I was 15 years into my career, and I knew that I was failing badly. I'd give it my best shot, and he would give me that little shake of the head again. He'd kind of sigh, and shake his head and look down.

Imagine how Immortan Joe felt in Mad Max: Fury Road, when the British guy wants a shot, and goes in and completely screws it up, and Immortan Joe goes, "mediocre" — which everybody who saw that scene apparently misunderstood. Mediocre was a high compliment from him, because that performance was not mediocre, it was terrible. Mediocre means kind of okay, lukewarm, tepid, middle of the pack, not awful. That's what this interviewer was doing every time I tried to answer one of his questions. And I was getting more and more frazzled. I maintained some humor about the situation, but I failed it badly.

So on the way home after my crushing defeat — it was like a video game where you're fighting all these weak boss monsters and you're like, what kind of stupid game is this? And then you're the final boss monster. Did you ever read Ender's Game? At the end, when they see the final boss, they're like, this is stupid, nobody prepped me for this. Honestly, it was as if they had put me in a room with John von Neumann asking me computer questions. So of course I failed, and I was sad. I was sad that I had to go back to my miserable existence of having lunch with Jeff Bezos and everything.

So I was driving back to my place in Seattle over the bridge, and the worst thing happened. It hit me. I realized what the answer was to his question — the 10x dumbed-down one. Because we were down to the hundred-x dumbed-down one before I finally got something right. He just said, all with a shake of his head and a sad look in his eye, this question that you're about to solve now is a hundred times easier than the one I first asked you, so you better get it. And I got through the hundred-x dumber one. And then on the bridge on the way back, I realized what the 10x dumber one was. I was only half an hour out of the office. Of course I didn't have access to any computers. This was before the iPhone was big, because it only came out the year before, so I didn't cheat or anything. I came up with the answer and I was like, oh no.

So I got home, and as fast as I could, I sent it to the recruiter and said, I realized what the answer is while I was driving. I'm going to code it. And I coded it up in C and sent it off to her, working. And nothing. Didn't hear anything.

I got a chance to ask him years later. I said, hey, after my interview, I realized what the answer was, and I sent a working implementation — did you see that at all? Did that matter at all? And guess what look he gave me? It was like I was that little kid asking if he could fly up to the top of the building. He kind of shook his head and looked at me sadly. And then he offered, he said, it might've shown some motivation.

His mannerisms are so much like Bill Belichick. Just think how tired Bill Belichick is. He loves his job, but how tired he is of idiots around him. That's how this guy was. And even after I joined Google, I was still terrified of him. He's nice, a great guy. But terrifying — existentially terrifying. When you spend your whole life being just a little bit smarter than your classmates, then you hang your hat on that, you get your identity tied up.

Going to Google was very much like that. People talk about imposter syndrome, how you show up there and you're so intimidated and you feel like you don't belong, and it's because the person who wrote your economics textbook is sitting at a desk next to you. The things people are doing there, and their credentials and their pedigrees — they literally got the smartest people from all over the world to come work at Google. So it was super intimidating. We would always tell stories about how intimidated we were. But the recurring nightmare that happened for me —

I've only had a few recurring nightmares in my life. My first recurring nightmare was that I had to go back into the Navy, and that went on for 10 years after I got out. I'd dream that I would just up and reenlist, and then I would come to my senses when I was out at sea on a boat, going, oh my God, what about my job and my life. And then, maybe five years after that, I got introduced to a new recurring nightmare. And it was about my Google interview.

I guess I should finish my interview story first. They liked me enough to send me off to a second round, even though I had completely tanked on my third interview, because Tom went on my website and read about my game and read my docs and decided that was a better signal than his interview. And he changed his recommendation that I could get a second round. And I went down to Mountain View for my second round of interviews and I wound up getting fours in all three of them. So they decided that was enough to bring me onto Google. And I did Grok there. So I feel like they weren't mistaken to hire me, even though I kind of didn't belong, because everyone there is crazy smart. I enjoyed my experience there, and I'm grateful to Tom for having changed his vote so that I could get another shot.

But I did get a chance to tell him about my recurring nightmare. Because the nightmare was about him. The nightmare was, I'd be working along at Google one day, and I would get a note from HR, and they'd be like, oh, Steve, we're so sorry, this is a big screw-up on our part, but you're going to have to interview again. It's just a formality, don't even worry about it. We lost your paperwork and we've got to interview you again. It's all going to be softball questions. And I'm like, oh, okay, fine. So I go down to re-interview.

I had this dream at least 20, 25 times over a period of a couple of years, and it was almost always exactly the same, unlike my Navy dreams, which were always slightly different. I'd go down to Mountain View and talk to a couple of people and they'd be like, oh hey Steve, how's it going, sorry we have to do this, can you just show us how to reverse a linked list or something? And I would do it. And then I'd be on the last interviewer, and Tom would walk in, and he would say, hey, do you guys mind if I just listen in, kind of audit quietly.

And in my recurring nightmare, the Googler who was re-interviewing me was as terrified of Tom as I was. And the interviewer asks me a question, and Tom calmly and politely says, if you don't mind, I'd like to ask a small variant of that question. And the interviewer's like, oh, by all means. And Tom gets up to this giant whiteboard in the interview room and he starts writing in Egyptian hieroglyphics. This is how my dream goes — it's always hieroglyphics. But it's math, it's his mathematical notation for some problem he's been working on that's related to the problem.

And in my fevered nightmare, I would break into a sweat. I would know the question he was asking me. I'd be just clued in enough to be able to say, I know this question, and I could never answer it in a million years. If you ever watched the movie Amadeus — it's about the life of Mozart, a bit over-dramatized, but a wonderful movie. In that movie, Salieri, a rival court composer who's nowhere near as good as Mozart — his curse is that he's actually good enough of a composer to understand how magnificent Mozart's music is, more than other people can. So it makes him feel doubly crushed that he's that bad. So in my dream, I look at this question, and he says, would you mind solving this for me? And I wake up in a panic, in a cold sweat. I had that nightmare probably 20, 25 times.

And I got to tell Tom about it. We were hanging out on a company trip, and we were having a drink, and I was emboldened by the alcohol, and I decided to tell him about my recurring nightmare about him. And he listened, and he laughed — a hearty belly laugh. And as we were walking back, he said he was very, very glad they hired me. So I was very pleased that he brought closure to that. And I probably only had the nightmare seven more times after that. So I think he managed to cure me.

So anyway, folks, that's the story of my Google interview. I guess I'm running long on time, so I'll have to tell you about my first week at Google some other time. I hope that if people figure out who Tom was, nobody's offended — most of all, Tom. It's important for us to share that just because you got into Google doesn't mean you waltzed into Google. People who get into Google often feel justifiably terrified that they made it in at all, and they feel justifiably small in that kind of environment. It was definitely one of the weirdest interview experiences I've ever had on either side of the table. And it did wind up with me having a recurring nightmare about it for several years afterwards. So I thought it was a good story. All right, folks, that's it for today's Stevey's Tech Talk. Tune in next time.