Here's something I've been playing around with off and on for months, and I think it might be sort-of ready for prime time. It's a demonstration of how evolution is not random chance, by recreating Richard Hardison's program that Richard Dawkins demoed so many times.
Start with a phrase (by default, "To be or not to be"), and click "Start!" It will then start a race between a completely random "monkey" and one that uses mutation and natural selection. Unlike Hardison's algorithm, which simply kept the correct letters and changed the rest, I give the incorrect letters an 80% chance to mutate, and the correct letters a .5% chance of mutating (a deleterious mutation).
The monkey CAN win. If you put in a one-character fitness phrase, the monkey should win half the time. He wins sometimes with two characters, occasionally with three (only once in my experience), and of course the proper phrases are so long it'll never win. So, when creationists talk about these big long genes being so improbable to form by chance, lay this little gem on them. AGREE with them that it's practically impossible by chance, and show them how evolution is NOT chance, but works much, much better, evolving a phrase in a few seconds that would take greater than the lifetime of the universe to achieve randomly.
Here's the url: http://www.shanekillian.org/evolve/evolver.html
I call it the "Really Too Simple Evolver." The first reason is that this isn't a proper genetic algorithm; there's not an actual generation going on, just modifying the phrase. This is supposed to simulate a population, not an individual. The second is, of course, that evolution is not goal-oriented, and the third is that in the real world there may be many different ways of achieving fitness whereas in this simulation there's just one.
But it's enough to give you the idea.
Well, "what do you think, sirs?"