This is a variation on the Weasel program.

You have a population of 10000 individuals.

On each generation:

• The individuals are each measured for their closeness to the phrase "Methinks It Is Like A Weasel". The closeness algorithm is to calculate the difference between each character in the individual and the corresponding character in the target phrase. These differences are used as a vector to calculate the scalar distance to the target. This is then cubed to make the target much more favourable.

• A new generation is created by "mating" individuals with each other to produce a new individual.

  "Mating" is:

  • Choose two individuals at random, but with a likelihood of being chosen in proportion to their closeness to the ideal phrase.
  • Crossover: for a random value 0 < n < individual.len, take the first n chars from the first individual and the remaining chars from the second individual.
  • Mutation: if some random threshold is hit, change one of the letters in the new individual at random.

At present the population size is set to 10000 and the mutation chance is 1%. The alphabet is all upper- and lower-case latin chars and space.

I haven’t done any concurrency yet, but shouldn’t be hard. Still, it looks like it does over 100 generations per second on my Mac in --release mode.