Recent work on Self-Replication in CAs

In the last 10 years or so, a lot of people have published work on CA machines that are capable of self-replication, but are much smaller than von Neumann's Universal Constructor, e.g.

But!

These simple machines are not universal constructors, and (more seriously) they ignore many of the theoretical considerations about evolvability that went into von Neumann's Universal Constructor.

McMullin discussed the shortcomings of Langton's work at some length in his PhD thesis (1992)