I don't know, were coding tests as implemented ever really appropriate? I work on 90th percentile theoretical stuff, and there's maybe three times in my career that I sat down and coded an algorithmic puzzle from first principles without a reference solution to work off of.