Re: OpenBSD for the win @iEgoPad
AIX was derived from AT&T code
AIX was derived from all sorts of things. I was working at IBM in the early AIX years, and I've seen source code for a number of its incarnations.
AIX 1 for the RT PC was largely SVR1 / SVR2 based, but IBM and ISC also incorporated a fair bit of BSD userland code. And that version of AIX ran on top of the RT's VRM microkernel, so it was hardly stock System V.
AIX 1 for the PS/2 was a different codebase, written by Locus under contract to IBM. Also SysV based. Locus also did the initial AIX/370.
AIX 2 for the RT continued the mix of SysV and BSD code, but much of the kernel was written in PL/I. (I don't know if that was true of the AIX 1 RT kernel as well.) Obviously those parts had little to do with AT&T code.
AIX 3 was a major rewrite, incorporating parts of SVR4 and BSD 4.3, and dropping the VRM in favor of a new pageable, modular, dynamically extendable kernel. Features like the unified VMM, logical cross-volume partitioning, JFS, and XCOFF executable format made AIX quite different from SVR4 at the system level, while SMIT, the ODM, and the like gave it a unique userland for sysadmins.
Subsequent versions diverged further from SVR4 in some ways, but - largely under the auspices of POSIX, the Austin Group, and ultimately the Single UNIX Specification - AIX userland and APIs gradually converged with other UNIX implementations, so it eventually got things like dlopen, POSIX threading, etc.