Well phooey! I’m still 150 comments behind … I’ll just throw this in, feel free to ignore.
Eric didn’t seem bothered when I pointed out where Dembski CSI (Dembski 2005) requires that probabilities can be greater than 1.0, and are even used in an example where CSI is negative. If violating the very definition of probability isn’t a refutation, then what is?
The WEASEL program finds the target by evaluating the function “number of matched letters to [TARGET]”. In that example the input target was hard-coded. BUT it could easily be otherwise, The algorithm can be applied to any input function (which might be input from a source external to the program. In term of evolution, the input comes from the environment, which is not coded in the DNA.
I recall we have a discussion of this last year. Among the many objections to Eric’s argument was a lack of definition of what was being computed, what constitutes a Halting State, and why this should apply to evolution? (A point I never got to make in that discussion was that the only relevant halting state seems to be “extinction of all life on Earth.”)
An excellent example of a Genetic Algorithm conquering a fitness landscape. 
I also recommend, Genetic Cars, inspired by BoxCar2D. It has fewer options, but runs multiple car in parallel, meaning you don’t have to watch it for nearly so long to see the progress.
The fitness landscapes for BoxCar2D are separate from the Genetic Algorithm, NOT hard coded as they were in the WEASEL example. BoxCar2D has a literal “landscape” that the cars must traverse, and the only input “fitness” is the distance traversed.
OK, now someone can tell me my late comments are now irrelevant.