The technical interview practice gap

You can really leave a lot of money on the table with this attitude of "oh well, I dodged a bullet, if the interview was hard then the job is probably bad and the whole company too."

Do executives have to demonstrate their knowledge of acquisitions or something similar?

Also let’s not act like this only applies to FAANG. Every dev job from internships and 50k entry level positions and up require the same testing. How many 70k/year jobs require testing every time you change positions outside of tech? Very few. Law, some medical fields, and a few other highly skilled fields do require licensing, but you only do that once per state at most.

Tech is the only field I’ve ever seen where your experience doesn’t matter and you have to prove your skills with every interview. In other industries, it’s mostly about personality.

Plenty of easier process exist but they are not jobs you can apply for. A baseball special assistant hired by an owner can be hired after drinks.

In terms of regular jobs a sales staff working on commission could be easier.

Tech hires quickly and frequently for the short-term. The algorithms test approach is intended to condense what would be a much more expansive investigation into something straightforward (and supposedly quantifiable.) This discards a significant amount of information which is considered in other high-end jobs.

I can understand the arguments for many of those, but... uh... somehow you're going to need to prove yourself. We can't just throw everyone who shows up into the position to see how they'd do... and the HN gestalt would presumably complain about how awful that is, too.

The mindset during consulting recruiting seems to be more of "we want to make sure you have a decent base level of knowledge and problem solving skills, but if you don't know how to code/make a PPT/whatever, that's fine, because we can teach you that stuff". Big tech recruiting seems to be the opposite: even when big tech companies claim "you don't need to know everything, because we'll teach you", they still require you to jump through these leetcode-style hoops to prove your skills.

And this is echoed beyond the hiring process too, in my experience. Consulting companies hire smart people and want to keep them for a long time, so they invest a lot of money into training those people and attempting to keep attrition down. FAANG on the other hand seems to care less about attrition (sometimes even embracing it, eg Amazon) and wants to hire people that require minimal training, and thus wants people that have immediately demonstrable knowledge so they can be immediately productive.

Interviews are significantly easier as far as prep time goes for doctors & psychologists.

BUT, you don't have to go through the entire med school & residency rigamarole as a software engineer either, so as far as total effort goes FANG is probably easier. I posit that software engineering is probably the most egalitarian method of socioeconomic mobility, because the only real barrier is time, a laptop, the internet and motivation, while all the other methods require an upper-middle class background and expensive schooling (law, medicine, finance) or just plain luck (business, celebrity art & sports)

But being an interviewer in FANG, I've actually found it rare to find the self taught type in practice. But we do leave that door open unlike medicine.

The key difference between our professional experiences is that the legal profession is self-regulating, whereas software devs/qa/pms are totally socially atomized.

Lawyers have their own organizations, which give them some independence from their clients. Society accepts that this is desirable for the maintenance of the legal order.

Software occupations should move in this direction, for the good of practitioners (wages, working conditions, fairness in promotions and hiring) and the wider society (complement to government regulation around security, privacy, fairness in hiring and promotion).

The bar association defines the examination that lawyers must pass to practice. Employer/client interviews focus more on "culture fit" and familiarity with specific practice areas than on certifying basic competence. The bar also administers "continuing education" to ensure that a lawyer who passed the bar decades ago isn't totally out of the loop on recent developments.

I would love to see the software sector move toward this kind of self-regulation through unionization. We have the power, we just need to coordinate.

Instead of a high-stakes and all-too-often arbitrary interview every single time you want to change jobs, you could prove your basic competence to practice through an exam, and elect to fulfill your continuing education requirements however makes sense for you, within a framework decided upon by workers themselves.

Such a system would not be perfect. No set of examinations is ever going to perfectly capture ever-shifting real world requirements, for example. But we can hardly do worse that the sector is doing now.

We have to stop letting recruiters (i.e. middlemen) and employers (i.e. counter-parties in the employment relation) dictate terms to software workers arbitrarily. The wider society should support tech worker organization in unions because these organizations would provide a check on executive and investor power. Without such checks and balances, we can only expect more of the same short-term thinking and ethical corner cutting.

I sure don’t think leetcode interviews are great, but I’ll take it compared to that.

Doctors in particular (with residency) and lawers to a lesser degree (bar exams) have a more difficult path into the profession. But they don't have to prep and go through the bar exam for every single interview after that.

It would be unthinkable to make an established surgeon balance chemical reaction equations or something obscure like that from their undergrad days as part of the interview! But that's what software engineers must deal with.

Doctors have board exams. Lawyers have bar exams. Financial professionals have their own set of exams. Electrical and mechanical engineers have professional engineer exams.

The difference is that they're standardized, so the prep industry caters to passing those exams more so than passing the interviews.

Software engineering is far from standardized because the subject matter is far too varied, so instead companies handle their own type of exams customized to their needs.

There is also a cottage industry around interview prep for lawyers, doctors, and other professions. It's just not well known because it's extremely expensive relative to spending a couple hundred bucks (if that) on Leetcode. It often takes the form of expensive 1:1 coaching that starts in the hundreds of dollars and can easily enter five figures. Students from wealthy families frequently have private coaches for this.

There's a lot luck involved because there are so many candidates, most of whom cannot come up with an approach to solve a problem (let alone to solve it well enough to serve billions of users well). So even most candidates who are qualified can't convincingly demonstrate that given the time constraints and pressure etc.

It is hard to get into FANG for college new grads, it is even harder if you're trying to break into the tech industry from outside without the schedule flexibility and focused community support of a top tier university. It is possible but you need a combination of high ability, strong drive, and luck.

At my FANG, we've made some interviewing changes over the past 5 years such that these days you'd spend at least 20% of your day interviewing talking about what you worked on and achieved and how you solved problems, whereas in the past it would be entirely at your interviewer's discretion and typically only be a few minutes of unscored small talk.

FANG is an interesting life experience and it pays well, but plenty of people find it is not all it is cracked up to be. Just as in every job, a lot depends on the particular people you work with and whether you enjoy the particular problems you are solving. So maybe also consider whether you really need to go FANG (right now) or whether you could be just as happy at a place that's willing to take a more individual approach on hiring. That place could even be a stepping stone to reach FANG later on if that's what you want to do later.

You can either dedicate the time and practice, or decide you don't want to do it since your situation does not enable you to do so (e.g. due to kids, family, or simply lack of motivation).

Of course, there are people who do not have to study. It seems you are not one of them. (Neither am I to be honest, this is just an observation, not an offensive remark). What else is there to do other than studying or not passing the interview in such a case?

Furthermore, the process is partly about how good you are at being able to learn, and partly about how well you are able to keep yourself motivated to grind something out, or a combination of the two. Given enough time, FAANG interviews are 100% something you can study for, and furthermore, it is expected that you spend the time to study for them as that is also a signal of something in itself.

On the job, you'll either have to be smart enough to pick stuff up quickly or spend the time to grind out the work (working overtime, etc). You also can't fuck things up too much because the fallout can be really expensive, so it helps to be able to do things right the first time. I think what is generally tested in the interview correlates with these requirements, and prevents false positives (and probably lots of false negatives too but you can just study more and try again at that point).

I am oversimplifying it, but the same forces that created the FAANGs can allow you to avoid them and companies like them, avoid SV. The world has become more evenly distributed in some ways, but HN and channels like it can make you think you have to pass through these certain gates.

What difficulty of questions will be asked? Easy? Hard? Do people actually want to "hear the thinking process" or do they just want an optimum answer regurgitated quickly? Do I need to practice "regular" interview questions also ("What's your greatest weakness?")?

All of this with a life-changing, potential millionaire amount of money at stake, which isn't great for the nerves.

Still, the vast majority of applicants aren’t going to pass. It’s not like it’s just some minor brushing up for most - it takes real work and dedication, for some that could do the job well over 100 hours... an infinite amount of time probably won’t be enough for those who can’t.

This is not a specific judgement about diversity as a metric, but an observation of making any something a metric.

We had a hypothesis internally that some parts of the interview were redundant. The interview was certainly tweaked over time, but the data science team found otherwise. All scores are positively correlated. And removing any major part of our interview would have made the interview worse. Given all my experience with it, I believe the triplebyte interview to be a fair assessment of candidates, and respectful to everyone involved. (Though I’d be very interested to know how much 5+ practice interviews would move the needle on scores, as this article claims.)

The hardest part of my job was nervous candidates. Stress is designed to help us survive running away from a lion. When we get stressed, we shut down our ability for creative thought and our memory gets weaker - which is literally the opposite of what you want in a job interview. Most of my energy as an interviewer went into helping candidates stay calm throughout the process, and I wasn’t always successful.

I don’t think there’s anything fundamentally wrong with whiteboard interviews. You’re right - we could probably rearrange the professional world to avoid all stressful assessments but I tend to agree with this article. Just do a bunch of them and you’ll get used to them. Being interviewed is just talking for an hour or two with someone smart about something you care about. Usually you get to talk to someone you would have no access to in your daily life. If you can get past your terror of being judged, they can be a lot of fun.

So a candidate a invests 60 minutes of their time and a multi million dollar company invests 10 minutes.

We should all boycott these asymmetrical "take home" tests.

Literally this. You can pretty much do all of the interviews required for FANG without talking to a human being, which from the FANG perspective, is absolutely GREAT, because it takes all of the existing social biases (e.g preferring a certain race, preferring a certain gender, etc.) out of the window and focuses ONLY on the algorithmic knowledge of the applicant.

Of course, there might be some interview here and there about soft questions, but for the most part, they don't really give a shit about that, and it's unlikely that you'd be rejected based on that, unless you _really_ suck at bull-shitting success/failure stories (which I am - can't stand doing it).

I think the last time I applied to Amazon (over a year ago at this point) I didn't even get phone screened. It was just a link to 2 graph algorithm problems that were the equivalent of a medium difficultly on Leetcode, and I was immediately graded. It's so fricking easy for FANG nowadays.

When your doing whiteboard interviews the interviewer is there the entire time interacting with you, so it's taking up 1.5hrs of engineer time at least for every 1hr interacting with the interviewee. Reviewing a take home takes less time.

but they're not? every screen i've been through has been on something that doesn't even compile the code (e.g. coderpad.io for FB or docs at G) and the interviewer then writes a summary afterwards (ala hiring packet at G) with ratings along some axes.

Take-home problems are automatically scored as well. The company defines an API or interface to the code and creates a test suite to run against it.

The benefit of take-home problems is that you have the option of reviewing the code in-depth if you need more perspective. If a candidate only passes 7/10 tests, I'll still dive into the code to see what went wrong. If it's an honest mistake, I ignore the failure and amend the instructions to make it more clear for the next candidate.

Students sort companies they are interested in in descending order so your take-home will get done whenever they have some time available.

With enough push back, I do think it's possible that those in charge of hiring in the world might reconsider how valuable leetcode tests are to judging potential employees.

Mostly because when I was interviewed the company liked me enough to skip half the process, including their code challenge stuff.

And that’s with being an immigrant from eastern-ish europe.

But it sure is fun to flog on it as a meme. I doubt more than 30% of engineers working in SV startups ever actually used leetcode. And I certainly wouldn’t hire anyone above absolute junior based on their leeting skillz.

[1] https://kstatic.googleusercontent.com/files/25badfc6b6d1b33f...

And with that said, you too should care, because I guarantee it's only a matter of time before those "hum-drum industries" take a look at the way SV does things and says "hey, we should do that too". This is already happening in many places. Your company could be next. You should be interested, if only to learn to spot the warning signs and nip it in the bud when you can.