I'm very interested in Colapinto's work. His dissertation, Articulating Space, has a lot of neat ideas and applications in it (http://versor.mat.ucsb.edu/ArticulatingSpace.pdf). I especially appreciate his explicit emphasis on exploration encouraging others to build on past work and discover what these formulations enable.

I'm curious in what way versor is 'more modern'. My understanding is that its main benefit over Gaigen is that there's no code-generation step, and provides more general library for various kinds of I/O. I have been tempted to use versor, but for reasons explained below I'm leaning toward Gaigen for optimized implementation.

What is attractive about Gaigen (2) is that it will generate code for Java, C++, and C# from the same specification. I'm working on a relatively high-level implementation in Clojure, and the possibility of compiling down to something which will run efficiently 'everywhere' is appealing. In particular, there's huge benefit of being able to remain within a highly interactive, dynamic environment for prototyping and experimentation. If I can work out the kinks of interfacing with Java there, then it should be relatively feasible to target C# (for Unity) and C++ for Unreal Engine (I'm especially focused on VR).

Also, the Gaigen authors literally wrote the book, so I suspect the least impedance mismatch in trying to work from their generated code. That said, versor is also obviously heavily influenced, so I wouldn't anticipate problems there.

At one point, earlier in my exploration, I had concluded that versor would definitely provide the most bang-for-buck. I've shifted, but I think that may be more a matter of my goals and requirements. I'm interested in any direct experience you have with versor or other GA software.