Having worked with Go for about a decade now, I largely agree that nil is a pain in the ass to work with, and the language has largely done nothing to make it better. However, Go (mostly) doesn't have exceptions. Ordinary problems are represented by non-nil errors, which are values. Panics exist but really are reserved for exceptional situations (with the number 1 cause being, of course, dereferencing nil).
There are a lot of ways in which Go handles nil better than C handles NULL. At the very least, a panic is better than a segfault. And carefully written code can avoid most kinds of nil panics entirely. So I guess the language's authors thought this would be enough to overcome the mistake. But I don't think they went far enough. The very limited type system and the lack of nil-safe operators make it not very ergonomic to write and read such "carefully written" code; design decisions in some key parts of the standard library completely undermine the language's attempts to minimize nil; and then there's the "untyped nil" default value for interfaces, which panics if you just look at it funny.
