I've frozen in interviews and been unable to think on simple problems just due to anxiety (which I've never experienced at work; if I can't think, I just say, "let me think about it" and I get back to the person in a bit). Other interviews I've completely demolished - solving problems that were supposed to take 30 minutes in 2 minutes, for example.

I think the type of problem and structure we use for interviews is bad. Let the candidate code alone without you watching. Give them internet access. Come back in an hour or two, or let them do it remotely or as a take home. This is more like our real jobs; and gets rid of cutesy "do you know this algorithm?" questions, which many whiteboarding problems are. Instead of going for pure data structure / dp / search / sort algorithms, have them implement things like business rules.

I'm 100% sure I've come off in a couple interviews like one of those "phew we dodged a bullet there, guy can't code at all" stories. Meanwhile actually I can and many employers have been very happy with my ability to do so, and I've had a long and reasonably successful career.

The problem's not even that I'm bad under pressure, and in fact I've repeatedly been told the exact opposite by people who've worked with me. The problem is specifically about doing a programming performance in front of an audience, in an interview rather than co-worker context. I'm god-awful at that unless I'm heavily prepped for exactly what's going on. No tools? I'll forget basic syntax, yes even in "a language I chose". Let me use tools? I'll forget how to use them, or get self-conscious and avoid things I'm not entirely sure will work, overriding my own muscle memory. It's a very specific problem but I doubt I'm the only one who absolutely can do the job just fine, but is also entirely capable of coming off like a complete "faker" in an interview.

[EDIT] to make matters worse I can talk about programming just fine in that setting, which probably adds to the "he's some kind of social engineering genius who learned to sound exactly like a competent programmer while somehow also not learning a single thing about programming" impression. It's not usually an issue, but I'm quite sure I've reinforced some interviewers' notion that it's a good thing they do whiteboard screening because they're overwhelmed with lying applicants, and that I was one of them.

Exactly. Yet if you listen to HN you get the impression 80% of the industry can't even do their job. Yet, if that were the case how the hell are they hiding that from their current employer??????

I've met one developer who couldn't do their job, and he moved into a pre-sales role shortly thereafter.

Everyone else was fine and did a good job but send them unprepared to a random interview on any given day of the week and I'm sure it'd be another one of those "well we dodged a bullet there".

The signal to noise ratio is Garbage.

> The problem's not even that I'm bad under pressure, and in fact I've repeatedly been told the exact opposite by people who've worked with me. The problem is specifically about doing a programming performance in front of an audience

Sure a "whiteboard interview" basically optimizes for people who can actually pass whiteboard interviews rather than people who can do the job.

But to claim that there is a NON-NEGLIGIBLE number of people who basically become 100% ignorant when under the stress of a whiteboard interview BUT are excellent employees in any other stressful situation is, to say it in nice words, [requires citation].

And if one really has this apparent rather unique type of handicap, then better mention on CV to try to get some empathy from the hirer, or just _train_ to avoid it. Yes, most people do _train_ for the express purpose of passing interviews.

I suspect some significant percentage of applicants may not have a disability, exactly, or always totally flub an interview in this way, but may do so often enough that it looks like the level of competence in the industry is much lower than it actually is, if one is taking one's personal experience with interviewees as an accurate measure of that. Add to this that assuredly some of the people confidently complaining about how 90% of their applicants can't write a for loop overlap with the ones generating complaints from applicants that some interviewers are themselves incompetent and asking broken questions (I guarantee you the people doing this think they're great interviewers and getting nothing but signal from their process, and they're probably also likely to exaggerate stories when relating them), then factor in a real tendency to take a 90% OK-to-good signal in interviews and practically forget it happened, taking the 10% bad as more accurate, and I think it's highly plausible the state of things isn't nearly as bad as some believe it is.

Tech interviews are remarkably scattershot in the form they take (outside well-known big companies), and are unusually anxiety-generating, even in the notoriously anxiety-filled field of interviews. Describe what one might (emphasis on might, part of the problem is that it's so often a surprise) expect in a tech interview process to some people outside the industry, and gauge their reactions. I definitely think it's likely they have even worse signal-to-noise ratio than is commonly thought.

[EDIT] Certainly I find it far less plausible that there's an absolute army of people out there, dwarfing the count of actually capable programmers, who are brilliant con-persons but too dumb to figure out that that skill itself is more valuable than programming, outside the top couple percent of programming jobs by comp, and apply it more directly to business roles that actively want it.

I don't think it's 90%. I'm saying there is a very, very low bar, but I would say it's hit by 20%-30% of candidates that reached me, not that many. Obviously this depends on much you offer, whether HR prefilters, and the like. But this is not 90% of candidates, I think.

And I'm quite confident it's not the interviewer. Usually the panel is quite anonymous when it comes to calling a "bullshit" candidate. We ask panel persons individually to avoid the "no one wants to contradict someone calling X bullshit'" effect.

I don't think tech interviews are _any_ worse than the ones outside. If anything, we have less standardization than other industries. I work for a engineering company first (software second) and the engineer interviews are basically _manufactured by HR_ (not engineers).

I wholeheartedly agree. Note that my second example was exactly that type of interview - we leave the candidate to code alone, technically sometimes in the same room as us but sometimes in a separate room, but either way with only a few "check ins" every 15-45 minutes depending on how things go. They have internet access, a base file and console session open that can run the file, and an editor already set up (not necessarily their editor of choice, though we've had people install an editor at the start of this process).

In this scenario, having someone literally be unable, in 2 hours, to replicate the first example code of some library, when they have access to almost identical code in the docs, is a pretty good filter.

Everyone experiences stress on an interview, and thus may not be able to solve "do you know it" problems that he would have been perfectly solved after a better night's sleep. That is not the problem (we call them "happy idea" problems around here). I really dislike being asked this type of problems and I also dislike asking them for these reasons. Despite the fact the positions I've interviewed for you end implementing "pure data structures" much more than you end up implementing business rules (we build software for engineering: simulation, optimization, etc.). So imagine how low I set the bar between my colleagues.

The problem is when you apparently _forget_ the syntax for a function declaration in a programming language you claim N years of experience in your CURRENT position. This is just absurd. Our you suddenly get completely stuck with counting odd numbers on a sequence or the like. And this happens surprisingly frequently. I just cannot believe you can have this type of anxiety (I am extremely shy myself) outside your first or maybe second interview.

I think my daily driver language (Ruby) has become so ingrained in my mind that it's written out of muscle memory in Vim, and translating that direct connection from thought to Vim into a whiteboard might actually be challenging. You almost never write for loops in idiomatic Ruby anyway, you use an iterator chain.

Languages like C that have a single official for loop syntax are a bit easier to remember, in part because it's a syntax I have previously used on a whiteboard rather than exclusively in actual code.

So if a candidate can't write a loop on a whiteboard, maybe let them try in a non-IDE text editor, or have them glance at a formal grammar for the language as a refresher that won't give away the game to someone who can't actually code.

Even if your IDE would type your entire programs for you, you would still know the syntax of the language just because I am 100% sure you have actually read more code than you ever wrote. And do note that even in C there is not a single official loop syntax -- a while (get_next())-like thing is a perfectly valid loop in C. All of them valid answers -- I used the word "sequence" and not "C array" for a reason.

And if I start showing them an EBNF grammar for the language, now that's TWO languages that you will show you don't remember rather than one. I find this even more absurd; you don't remember anything of the syntax of the language you used at work for years, but you think you can rediscover it from the FORMAL language specification in a whiteboard?