Yes, the AIX code looks nothing like SYS V. It's been 20+ years since I did a lot of work in AIX, and most of my work was in networking, external pagers, and NFS, and even there you could see it was different (although much of the NFS was clearly the Sun reference code, one giveaway being the Sun copyrights in it :-) I always thought it was an interesting code base -- they seemed to get preemptability right from the start, for example. As it was explained to me, IBM did a full implementation from manuals of both the kernel and the commands. 

There were lots of little weirdnesses in the commands. mkdir -p, for example, would give you an error if the directory existed -- they got the creation of the tree right, but the error wrong. There were tons of these little gotchas in the commands and it's one thing that made NTP and Condor, for just two examples, a real chore on AIX.

I visited the now-closed IBM Palo Alto center in 1991, and they told me an interesting AIX story. Seems to that point, on the mainframes, AIX had run under VM. The native port was either starting or soon to start, and there was some question about channel programming -- mainly, if the people who really knew how it worked were still at IBM, or even still alive. I guess they worked it out, however ;-)