I'm a general contractor (Unicorn Homes Inc.) and I'm currently building a number of spec homes. We're going to need electrical work done in all the homes and I'm in the process of sub-contracting out that work. What I need from you, Joe Electrician, is for you to go wire the kitchen of one of these homes. Once you're done I will inspect your work and let you know if I'm interested in hiring you for more work.
So it's more like "What I need from you, Joe Electrician, is for you to go wire up this demo kitchen frame that I've set up. Once you're done I will inspect your work and let you know if I'm interested in hiring you to wire up the actual houses I'm building.".
It's usually not "build out our new feature", because spinup time is a thing, but some companies do seem to throw around "build our new one-off microservice". Stuff like "given this incoming data, set up intake, a storage layer, and a backend layer that applies formula X". If the data and the formula look business-relevant, it's at least enough to raise an eyebrow.
That said, I'm not sure how much I care, provided the interview is sincere and not some kind of bizarre scam to farm out work. My main concern is much more with scale. I totally see the merit of replacing 5 hours of whiteboard interviews with 5 hours of "build a toy project at home". I don't so much appreciate "take a week of your life and build a substantial project so we don't have to do any legwork".
What I've seen is something like "this is a startup in market X, whose product doesn't do Y but could, and they're asking candidates to write code that does Y, and sign over all rights to everything provided during the interview process".
Optimistically, it means that if whoever they hire will be told "hey, first assignment is Y, you've already made progress!" Pessimistically, it goes in the same category as "Whartonite seeks codemonkey" and "sign this NDA to interview" - it's a matter of ego and ambition exceeding skill.
- A short phone interview using a collaborative editing tool like CoderPad.
- An example project they can complete independently and turn in. We've designed this project to take about 2 hours to complete.
Almost universally, candidates choose the "take home" project option.
Then you've got it right.
It's good to see people admit whiteboard interviews are a crummy system, I think you can learn way more from X hours of "do a task on a real computer with the internet" than you can from the same X hours of "do algorithms work on a whiteboard".
The objection, for me at least, is to "do 10*X hours of work on a real computer". It's bizarrely common to massively scale up the time investment when moving to this sort of task. And the answer is so easy - just don't do that!
If software were like homebuilding, there would be legally enforced licensing.
Would it be so bad?