When Knuth made that prediction, there were probably tons of rough edges to programming. But now we have powerful abstractions that make it much simpler. I think his estimate makes more sense if you qualify it as someone who is able to investigate and devise a bugfix for anywhere in that great sea of abstractions.
