I too am supportive of using LeetCode-style easy problems. FizzBuzz used to be all the rage, but dummies have memorised that now. :)

For a long time, I used an easy-to-understand, but hard-to-be-perfect interview problem: Please implement the classic C-function atoi() [string to int] in any language. (The candidate cannot use a built-in to do the parse!)

The best candidates immediately dissect the problem and start to ask about edge cases. The worst ask no questions and very rarely do a good job, and never think about test cases.

As someone who broke into the industry without formal CS education, contacts, or family networks and struggled to speak confidently and coherently in interviews, leetcode didn’t do anything for me except stress me out that much further.

I’ve been able to stay comfortably employed and even climb the ranks in software dev for 7 years now, but had it not been for the company that first hired me eschewing the leetcode in favor of practical exercises in their interview, I wouldn’t have gotten in until a couple years later at minimum, assuming the resources required to continue studying held out (I was extremely cash-bare at that point, so absolutely not a given).

Now that I have years of experience under my belt it’s not as much of an issue — lots of places want to hire me regardless, but if I really had to I could drill leetcode for a couple months to prep for an interview. Back in the early days that wasn’t really practical, though.