I don't think he defined his terms very precisely in his document; but as my comment says, yes, there are two levels. There are people who can approve changes to the implementation of Go, and people who can make changes to the specification of Go.
There are 100+ people who can approve changes to the implementation; but there only a handful of people who can make changes to the language specification, and all of those people are at Google.
> the implementation of Go, and people who can make changes to the specification of Go
Should that be "the reference implementation of Go" ?
The specification mentions "implementation restriction" many times, so the Go spec writers seem to be encouraging other implementations. Besides gc, google also provide gccgo. If you download the gc source, change it a little so it still conforms to the spec, and publish, then that would be another implementation of Go. I believe there's an incomplete JS-based version of Go around as well, so if that's ever finished, that would be another implementation of Go.
There are 100+ people who can approve changes to the implementation; but there only a handful of people who can make changes to the language specification, and all of those people are at Google.