[evoneutron]

It would probably take me less time due to not having the pressure of someone looking, but it might still take me ~ 30min.

I think the general problem is that I don't deal with such low-level problems on a daily basis.

I can write multi-threaded Kafka consumer app that processes thousands events per second and applies some basic ETL on them, add end-to-end integration test for it in <8 hours.

But it may take me ~1/8th of that time to write a low-level (quick) sort algorithm.

[mlthoughts2018]

Interviewers are very bad at accounting for this.

For example, there is almost no engineering question on the planet that people straightforwardly solve from first principles in 30 minutes, unless it’s totally shallow based on memorization or childish rote practice — and has virtually zero relationship to how they would ever solve any problem at work.

But there are all kinds of insanely hard problems that people can give you beautiful solutions for in 10-20 hours, and involve an entirely different way of working than anything possibly representable in 30 minutes.

Engineering and math questions just require burn-in and rumination, tinkering, stop & get a cup of tea, take a walk, tinker more. They just do.

Seeing “how someone thinks” in an artificial 30-60 minute session only tests a completely separate and mostly unnecessary skill set for passing tests and memorizing things.

[koala_man]

>Seeing “how someone thinks” in an artificial 30-60 minute session only tests a completely separate and mostly unnecessary skill set

Totally agree. However, what's a good alternative? A company can't hire everyone for a trial week, and a candidate can't work a trial week for every company. It's really hard to design a hiring process that is efficient for both parties plus resistant to abuse.

[mlthoughts2018]

They can ask probing technical questions about past work experience, projects and education. Pretty much like how every other industry interviews people.

You are right that interviewers won’t get much more than an hour to judge a candidate. So why waste it on information that fundamentally can’t be helpful (like whiteboard trivia)?

Instead, first jump right into a domain where the candidate should already be comfortable and technical — their past work projects — and go from there, probing for more technical depth.

The idea that it somehow makes more sense to use limited time by wholesale discounting & ignoring the person’s previous work history in favor of having them write a sorting algorithm in CoderPad is totally indefensible.