it takes the first letter of each text-chunk added, throws in a random letter, and then an "e" (for good luck) and then makes an anagram out of it. there might be some other stuff in there, too, but i forget!

it sometimes comes up with some interesting things, but really i need to get a lot of junk words out of the dictionary list. (like, why does it like 'catv' so much... who put that in there?)

also, this occasionally fails altogether (takes too long to process or cant find an anagram [not enough vowels?]), then it chooses from a list of hard-coded phrases.

so there you have it. certainly not all that exciting, but now you can go back over the last couple hundred levels and see the weirdness.

that said, i am up to ideas on deriving new seed phrases. a collection of random words would be the first choice, but is there something more interesting? maybe random from several levels? some kinda pattern matching? how can we get the "flavor" of the last couple levels?