The fuck kind of programming language is "smalltalk"?

I was thinking the same thing. I mean, I just did a coding test for a potential job, and I know I did at least as good as, and likely better than this.

edit: just to prove to myself, I went ahead and wrote the program without looking things up. I'm self-taught so I feel pretty proud. It took about 25 mins, and it works!

It is very good test for the ability to research, I think. The amount of people who painstakingly went through some video tutorial on PHP and are now developers is insane. I'm sure there's place in the market for them (writing Wordpress themes/plugins, for example), but it's hard to find a programmer with ability to think these days. Not because people are more stupid, but because every other person is a programmer now.

You had a day or two [...] none of the last 5 candidates could even send in a solution that would run.

As harsh as this sounds, this test was doing its job. Assuming you're not hiring junior candidates, that is.

One day is enough to research XSLT enough to get the gist, and two is enough for a polished solution. And since we're just stripping tags, we're really just selecting for all the inner text, which is weird but not hard to do with the right selector expression. The task also selects for people that understand XML processing as programmatically manipulating a DOM, which is crucial to wrapping your head around more advanced tasks.

grey and gray are both in CSS. If you've worked with CSS long enough there's a chance you've seen both. :)

I would just write down the steps I would take, just some psudocode. It doesn't have to work, it just has to make sense in the style of the language you're talking about.

import random library  
import any GUI/display libraries required for the outcome desired

build array of integers [1..52] (or 0..51 if you're being fancy)
for loop 1..1000
       select random number A 1..52 (or 0..51 if you used that above)
       select random number B 1..52 (or 0..51 if you used that above)
       swap elements in the array A and B
pop first two elements from array
decode at display time what the two numbers represent in terms of playing cards

If the test requires more than that, then they're crazy. The syntax doesn't matter, just that you can logic yourself through the problem.
You can use the IDE, google, or whatever to fill in the specifics. If you wanted me to do that in literally any programming language, once the psudocode is done, you just spend an hour or so looking up the details.

Even the odd numbers stuff? I think interviewers account for nerves and being outside an IDE. You might draw a blank but would you be would be randomly adding things like these did?

for some reason I still needed to know programming on paper so much so that at the university, in class prgramming we had to do the exam on paper. every time. no matter if it was Java or C++...

I strangely enjoyed it, but it still was kinda weird.

Actually, I prefer this one: for(_=[];_<+!![]+""+[]*[]+[]*[];_++%+(!![]+!![])?[]:console.log(_));

This is why I prefer live interviews. I tell them they can use whatever tools they want, search for anything they want, there are no restrictions. All I ask is that they share their entire screen (if not in person) and try to "think out loud" as much as possible. I then time-box each step (usually 15m ea in a 1-hour interview).

I am most interested in HOW they solve the challenges I set out for them. Whether they complete it or not is usually irrelevant.

Edit: Lately, though - I warn against AI. I don't ban it, but every person that has tried to use AI in an interview has gone down in flames. AI simply cannot be trusted... and if you haven't learned that lesson, and you can't even tell when it's giving you bad information... yikes.

I jump between languages so much I can never remember the structure.

for item in items? Or item of items or items as item?

Best to just have the IDE auto complete it.

This is on the easier end of the scale to be sure, but as someone who's interviewed candidates with similar questions, it eliminates a surprising number of people...

My theory is that modern coding bootcamps stuff their students full of buzzwords instead of letting them learn the basics

When I interview people, I don't care how they get an answer, I want to see that they can get to the answer, ideally the correct one, but it doesn't matter if it's wrong. I want them to show me their problem solving skills and that they understand their own solution.

If you can read existing code and understand complexities you are already better than 80% of these hires.

That's like stage one where you filter out the obviously incompetent ones.

You wouldn't believe how many candidates with years of experience can't figure out those simple problems. Or even the super well known fizzbuzz.

It's insane, people will claim like 2-3 years of experience with Ansible, they can't even get a file copied. Couple years of Python, they don't understand async, generators and other pretty basic features.

People have always been lying a bit about their experience but it's getting way, way out of control.

But they are the same thing.

If your client has an HD picture button (like Boost) you need to click it to actually receive the proper image