From what I'm hearing, progress on the open source OpenGL implementation for R600 and upwards (in marketing speak, those are the cards with HD in the name) is going very well. Thanks go to AMD for employing a number of developers who are pushing this all forward.
If this continues, my days as a hobby driver hacker may be counted. I'm looking forward to that with slightly mixed feelings. After all, it's been a pretty interesting ride, starting way back when I first reverse engineered my trusty old Radeon 9700 and wrote the initial driver many years ago. On the other hand, as software goes, hardware drivers are not lasting. There are other, more universal projects that I've been eyeing for a while without ever having enough time to really sink into them. In particular, the state of development environments on Linux is frustrating me, and I want to do something about it - but this is material for a future posting.
For now, interesting work in R300-R500 land is still ahead (again, in marketing speak, those are Radeon 9600 up to X1000-somethings, as well as a random collection of integrated graphics parts - marketing names for those are even worse than the marketing names for the normal, discrete graphics parts). In particular, Corbin recently merged my work on the shader compiler for Gallium, and I've been fixing random bugs here and there in the classic Mesa driver.
There are obviously bugs left - it never ends - and I've spent some time trying to isolate some particularly tricky ones, always keeping in mind that eventually, corresponding regression tests should be added to Piglit so that other drivers (and our future selves) will benefit.