Hi again Jack,
FYI, I've implemented a genetic algorithm in Automatic Investor 2.0 that finds the best (well, almost best) set of AIM parameters. I've added it to the brute-force algorithm that finds the *best* set of parameters (however it can take an unreasonably long time to do so).
This sounds similar to your evolutionary programming technique (although it's not identical).
I've also used a genetic algorithm to implement an MPT-based portfolio optimizer (based on work by Markowitz and Sharpe).
Seems to work very well.