Newbie here. I've been giving a-life cursory thoughts for a couple years now, but have never actually programmed anything. I'm reading Richard Dawkin's "The Selfish Gene", juxtaposed with Douglas Hoffstadter's "Metamagical Themas" and the spirit of the two put together got me to thinking about it more lately.

Here's my question:

With software a-life, how does one simulate random mutations/inversions/recombinations/etc. in the code for the a-life without putting too many restrictions on the possibilities in order to avoid errors. What I mean is that if the "genes" are just code, it seems you would have to regulate the nature of the mutations so as to avoid instructions that would mess up the flow of program execution, e.g. a mutation that in effect makes a JUMP instruction outside of the program causing all hell to break loose. Of course there are ways you could prevent this, but spelling out the "allowed" mutations seems contrary to the spirit of a-life. What kind of solutions are there for for this?