I think the author is saying we should do interviews like nearly every other industry.

Look at past experience, and ask the candidate to talk about that experience to verify that they aren't lying.

If someone has 5 years experience working at Google, and they can talk coherently about technical aspects of the projects they worked on, there is no reason to ask them to code on a whiteboard.

The way it currently works is this: I see you have a CS, degree and 10 years experience at a well known company. Excellent. Now you have 2 eggs and a 100 story skyscraper. Tell me the minimum number of egg drops you'll have to perform to discover the minimum height that will break the egg.

Every engineer no matter how experienced has to answer questions like they just left school yesterday, no other industry does this.

>Personally, I know that to be false. I have worked with incompetent engineers, who literally couldn't code FizzBuzz.

99 times out of 100 you could catch that engineer by talking to him about technical aspects of past projects he worked on.

"Look at past experience, and ask the candidate to talk about that experience to verify that they aren't lying. If someone has 5 years experience working at Google, and they can talk coherently about technical aspects of the projects they worked on, there is no reason to ask them to code on a whiteboard."

I always argue for this, and about 98% of the time people look at me like I'm sort of crazy man and our company is going to go down in flames. Everyone has an anecdote, "well, there was this one time ....", yet I'm convinced that there's roughly 0 real value add from all of the hoops and hurdles most interviews provide in terms of accurately identifying strong candidates.

I always try to make the argument that if someone can talk intelligently about a wide range of programming knowledge in an actual conversation - for instance, knowing that data type X would be a logical segue from the current point, and do that throughout the interview, personally I think the likelihood of them being an idiot are slim. However, most people I argue this to are always afraid of this, afraid that some interloper will sneak through unless we put them through a gauntlet of questions regarding things they'll never do in their day to day life.

This is why I like to give laptops to engineers and ask to make something simple, that might require logic, but doesn't require any sort of specialized knowledge. They can even look up documentation.

Like make a table view that has word counts for a string. Or make this simple puzzle game. Or make a simple address book app.

Just an FYI -- working on someone else's laptop can often be an unpleasant experience (weird keyboard configuration, different software installed, etc.)

My current opinion on this would be to offer a laptop so you aren't disadvantaging candidates who don't have one too badly, but let people use their own if they have one (and let them know ahead of time that it will be an option and what they should be set up for.)

We tell candidates up front that they'll be doing some live coding and welcome them to bring their own laptop if they would like. We have a backup around with just about every reasonable IDE and editor installed for our languages but even then I agree, it's definitely not the same using someone else's computer.

Wow... I understand that you'd be much more productive with a large monitor and good keyboard, but to the extent that you couldn't perform in a coding interview? Give me a break...

In the past, I've used CollabEdit for "phone screens" and the whiteboard for in-person interviews. I try to ask some "coding" questions (identify a better data structure / algo for a particular scenario and then implement it in code) and some problem-solving conversational questions. Seems to work pretty well, but I'm always on the lookout for better approaches. Part of the problem is that the software field is so broad that it's hard to get a sense of a person's abilities with such a limited amount of time to ask questions.

Sure, we'd all want an ideal setup but ideal isn't possible in an interview situation - both sides need to compromise a bit for practical reasons. And really, while we say "laptop" if someone wanted to lug in their desktop with two monitors and keyboard we wouldn't stop them. In fact, it'd probably earn them points for being extra nerdy ;)

Would you rather a laptop w/ standard editors and IDEs or would you prefer a google doc, collabedit or a whiteboard like many places do?

That's a bit like saying, "My daily driver is a Mercedes Benz SL55 AMG, so I can't drive a Hyundai Elantra, not even for a block or two." Or like saying, "Oh, I normally wear Air Jordans, so I can't walk with flip-flops."

Moreover, if you're that useless without your dual monitors and a full-size keyboard, would a whiteboard-based interview be any better?

One company I know of says bring your own laptop as a result, and another says they do a day of pair programming as the interview.

Hmm..This just pop into my mind as I read your comment.

Could you use this specific scenario to view how the candidate would react? If he starts to complain that the right tools are not on the laptop, or he is unfamiliar with environment, does it is signify that he is focusing too much on the tools rather than on solving the problem?

Maybe his time management skills needs some work, ie limited time to solve the problem and too much time worrying about environment.

Well, we don't do the puzzle questions, and I agree that those are not useful to determine the level of software engineering experience a candidate has.

However...

99 times out of 100 you could catch that engineer by talking to him about technical aspects of past projects he worked on.

That does not work for us.

We must put each and every candidate in front of a keyboard, and bang out at least a little code.

We had one candidate, recently graduated with an MS in CS. Nice personality, good talker in general. The candidate seemed to understand CS and could explain how to solve problems.

However, even with repeated prompting, I could not get this person to even type in a loop to start to solve the problem presented.

We've interviewed multiple candidates with a BS in CS or similar who also spectacularly failed at doing even basic coding tasks.

Some people are really great talkers. They really do sound like they know what they are doing. But that does not mean they can actually do anything.

As someone else said, I'm not talking about someone fresh out of school with a few months internship experience.

>That does not work for us.

How do you know, did you try it? I said 99/100, sure there are some rare people who may slip through, but there are people who slip through a coding challenge as well.

How do you know that your coding test is better at detecting secretly terrible engineers than talking through past projects (for experienced engineers)?

>Some people are really great talkers. They really do sound like they know what they are doing. But that does not mean they can actually do anything.

That's sounds like a problem with the questions you're asking. Some people can throw out buzzwords and talk at a high level, but if you really dig down into the implementation details of project you can weed those people out.

How is this relevant to the parent? You're talking about someone with a CS degree, which means they have no experience. How can you ask them about the technical aspects of what they did on a project. "What parts of the system did you write? And what problems did you hit. What solution or bug were you proud of solving?".

A degree is not experience, CS != programming.

Nearly all the the candidates that we've interviewed have at least a little job experience, such as an internship. And even for the ones that don't, they all have worked on team-based projects as part of their coursework.

So we end up asking them all those sorts of questions.

Working on team-based projects as part of coursework does not mean they've ever written a line of code in their life.

Going through a CS degree without learning to code is easy. Depending on your CS course you may hardly have touched any code beyond extremely entry level courses where you're being hand-held the entire way, if you pick the "right" subject.

CS is not software engineering.

Similar with internships.

For both those situations, I can understand that you want to verify ability to actually write code.

You're just agreeing with ansible - no one here is arguing that a CS degree means you can code. Rather ansible is relating his experiences with people graduating from those courses and not being able to code.

What you seem to be implying, is that this is obviously true for graduates, but couldn't possibly be true for people with real experience. (But your implication has no evidence to support it)

Exactly why do you believe that it is completely plausible for a recent CS graduate to be able to explain the technical details of their final year project when in reality they contributed zero code to that project, yet it is entirely implausible for someone with 3 years experience to be able to be able to explain the technical details of their most recent development projects when in reality they contributed zero code to those projects?

> yet it is entirely implausible for someone with 3 years experience to be able to be able to explain the technical details of their most recent development projects when in reality they contributed zero code to those projects?

Here is problem. It's not entirely implausible, just very unlikely.

The type of interview I'm talking about isn't going to catch every single bad candidate just the vast majority. But whiteboard interviews don't catch every candidate either and they have an extreme amount of false negatives.

To your point about students, they have much less work experience to talk about. The point of the interview is to look at their experience and talk about it to verity they aren't lying. In most cases the student doesn't have enough experience to tell me anything.

That being said, if a student can talk me through technical implementation details of a final project, then I think that's a pretty good indication that the student can code.

I would say the false positive rate for that test would be very close to the false positive rate for a white board interview with drastically lower false negatives.

If you're Google and you have a limitless number of people who want to work for you, you can get away with taking a vastly higher false negative rate for a slightly lower false positive rate. However, if you're almost every other company out there, you might want to reconsider.

> Going through a CS degree without learning to code is easy.

Let's name some schools here. In no school I am reasonably familiar with (either first hand or by friends with attendees) would this possible.

I am not sure if your comment is supposed to reply to mine.

I do think there are incompetent engineers, and I have interviewed about a dozen of them. It is very easy to filter them out.

The challenging part is to evaluate how good the candidates are, and if they would be good for the team and for the applied positions.

In start-ups with <20 engineers, you would like to have the candidates to be able to wear many hats. In a big engineering team, the requirements could be different. For example, besides challenging and creative work, there could be some tedious, even boring, work. The latter is also important, both for the whole system running smoothly and for the long term health of the system. So if the candidate is sufficiently competent, very detail-oriented, disciplined, team-oriented, she/he would be an excellent fit for the job, even if she/he is not technically very strong and couldn't figure out challenging technical issues. An excellent engineer, on the other hand, might get bore, grumpy, and leave in 6 months. It is expensive to restart the hiring process.

I worked with a guy who, after having been at the company for 6 months, called me over to his desk to help with a "problem." He wondered why he was getting a syntax error. Okay, fair enough. I looked at his screen, and he was putting the arguments to the function OUTSIDE the parentheses. The guy had clearly just been copying and pasting code until "it worked" the entire time he had been there. I almost palmed my face right off of my head.

To be fair, I have done that before when just messing around on my own late at night. After working for as long as I had been I felt like reading my own code was like wading through sludge and my eyes just didn't catch the error. I was still however well aware that putting the arguments outside the parentheses is not how this works, not how any of this works.

That reminds me of how I deal with C's pointer syntax. "Well, it's either * or & or a combination of a low number of the two, let's try a couple of combinations that seems vaguely reasonable... and some that don't, just in case". I don't know who's most at fault: Me or C.

In case it helps, I've noticed what I'd consider a wart in C that seems to be a common stumbling block with this - including when I first learned C a couple decades ago: * when used for types, has the "opposite" effect when it's used for values/expressions.

  int* foo; // The addition of a star here, on a type, means we're converting the type /into/ a pointer-to-int /from/ a plain int.

  int i = *foo; // The addition of a star here, on a value, means we're converting the type /from/ a pointer-to-int /into/ a plain int (reference).

I found it much easier to keep the two straight once I segregated their use cases like this. There's a similar duality with & in C++ using references, although not quite as cleanly:

  int& foo = ...; // The addition of & here, on a type, means we're converting the type /into/ a reference /from/ a plain int.

  p = &foo; // The addition of & here, on a value, means we're converting the type /from/ a reference /into/ a pointer-to-int.

Regardless of "fault", I hope this helps :)

(edit: C++ style comments used even in 'C' code to try and prevent YN from treating stars as formatting... replaced a few instances with 'a star'... changed emphasis to use slashes because it's still eating asterisks characters...)

rule of thumb: "*" means "look up the area of memory this value points to". "&" means the reverse - "give me a reference to this area of memory".

The problem with randomly trying things with pointers is that sometimes you can accidentally fix the issue but introduce a subtle bug that's dependent on the data being stored.

I'm with you.. I'd have to read a refresher on getting started with C if I were to even do something trivial with it... I've been in higher level languages so long (C#, JavaScript, etc)... I find it helpful to understand the concepts sometimes, but would be unlikely to reach for something more low-level than go these days.

I'm sure there are several variations but I have a sort of mnemonic to remember this - that * looks like an arrowhead seen head-on and so it "points" to a value. & is then, a reference.

I've been fortunate enough to work on small teams my entire career where someone like you describe simply would not last. I don't think I've come across a person this bad though. Even the 'bad' programmers I've worked with were able make working code, but it was not designed well.

>I have worked with incompetent engineers, who literally couldn't code FizzBuzz.

Operative word worked. This is exactly what the article says: "They always seem to be elsewhere. For everyone."

Uh, no? For the time when I was employed there, they weren't "elsewhere." They were sitting right next to me.

(The presence of incompetent engineers was certainly a factor in why I left that job.)