One big problem is that companies (especially FAANG) are wasting time having very senior devs do Leetcode-style shit, which is a false negative for people that don't balance binary trees and manually regex strings all day. This especially penalizes more experienced / older devs who don't have time to LC grind (of which I'm also one), and introduces unconscious selection bias.
Certainly there are other ways to verify someone can code: code samples, talking through high level language concepts, reviewing existing code, etc.
What people should be focusing on is this - 1) Do you understand enough systems architecture / code techniques to succeed at this particular company? This can be done via mostly talking. 2) Do you have experience that will help us with our company-specific problems? Give them an actual problem you regularly deal with, and get them to attack it.
Certainly there are other ways to verify someone can code: code samples, talking through high level language concepts, reviewing existing code, etc.
What people should be focusing on is this - 1) Do you understand enough systems architecture / code techniques to succeed at this particular company? This can be done via mostly talking. 2) Do you have experience that will help us with our company-specific problems? Give them an actual problem you regularly deal with, and get them to attack it.