John was and is an important figure in developing and promoting genetic programming, and came up with several of the common algorithms used, and holds several patents on the technique. BUT he was neither the inventor of GP nor of its most common form (s-expression tree-style -- what's being discussed in the blog). That honor goes to Nichael Cramer, who published the idea at ICGA in 1985.

http://homepages.sover.net/~nichael/nlc-publications/icga85/

The term "Genetic Programming" may have been coined by Hugo de Garis.

As to GP implementations. In the research world, there are probably four major ones. I'll list them in the order I think they're presently used:

1. ECJ. Java. [I wrote it, hehe].

2. OpenBEAGLE. C++.

3. EO. C++.

4. lil-gp. C. lil-gp is still quite serviceable but a bit long in the tooth now. I used lil-gp very heavily before writing ECJ, and in fact many things in ECJ were inspired by it.

It seems like it should be much easier to do this in Lisp. In other languages wouldn't you need to work on the AST instead of on the source code to have much chance of generating syntactically correct programs ?