The latest trend for tech interviews: Days of unpaid homework

The problem becomes, but then when would play videogames?

It's fine to be eager for an occasional puzzle challenge, but don't forget about the opportunity cost. They want 8 hours of your videogame time for something that likely has a low return on that time investment (ie, the homework assignment is [not] productive code; there's no guarantee you get the job as that homework assignment is only input in the process, etc).

That sort of opportunity cost favors the young (with no or small families, and a stronger ability to live on fewer hours of sleep), the eager (passionate for that particular job), and generally those with fewer work/life balance concerns (what if you need those 8 hours to stay healthy or protect your family?).

If you are already employed, you risk burning out by trying to burn the candle at both ends and do unpaid work in your off-hours. There's a possible opportunity cost that your day job's productivity suffers and you potentially lose your job.

Even when unemployed there are opportunity costs at play (if two companies need 8 hours in the next 12, which one do you turn down). The old adage is that unemployment is a full-time job (hunting for prospects, networking, talking to recruiters, dealing with unemployment wage systems, etc), and is stressful enough without adding additional unpaid labor on top of it.

Yes, bad english can be detrimental: starts wars, malpractice, etc... We just take it for granted.

If code is treated like math, then we can talk about absolutes and good code.

But 90% of us talk about it in the language context, hence why we call it a programming language....

(Being forced to use someone else's machine is a developer hell with so many dimensions.)

Do you want to work with team mates who have been vetted over less than 2 hours? I've worked with people who couldn't code, it's not a fun experience.

There are a lot of people who can talk a good talk but can't actually code. This was the reason for the FizzBuzz test (which I dislike, but it was invented to solve a real problem).

My current employer would not be able to do contract-to-hire or a trial contract for legal and policy reasons, and I strongly believe that a homework assignment taking approximately 2 hours, when combined with another 1.5 hours in person provides the best signal of all the methods I've tried. That's a total interview process of 3.5 hours i.e. about half a day. If you can find half a day for an in person interview then you can certainly find the same hours split up differently for homework + in person.

Trying to make it about sexism is crazy. Homework is much more flexible as to when you do it than in person interviews. Just because you're looking after kids doesn't mean you won't be able to take a couple of hours when they're asleep or in school. If doing homework is a problem, then I'm sure doing an inperson is even harder.

Our homework + inperson extension is on a topic related to our real field, not just a compsci quiz (which means it gives you a chance to show off domain knowledge if you want the extra points), allows no frameworks (except testing), produces an interesting artefact at the end, and can be completed in less than 100 lines of functional code plus a smattering of test code (although many will do more, sometimes crazily so - for which they lose marks). There are exhortations in the instructions to keep it simple. You can schedule when you want to work on it yourself, work on it with your own tools and hardware, split the time over a couple of weeks if it's difficult to get a block of time together.

The only real question is the time commitment. I've targeted 2 hours, but many spend around 3 or so, and some spend more on it. Setting an 8 hour challenge for my kind of work woud be entirely unfair to the candidate, unless that was the total time spent on the entire interview process.

I've hired with a number of different processes, and this one takes the least total time for the candidate and seems to have given me the best results.