The rules aren't quite consistent. Generally, the ball fits under a bridge raised by one full block or more, but for no apparent reason this doesn't apply to level 25. It bounced off the bridge that it fit under and forced me to raise it more.
I think that it ought to have a better way of judging whether a piece is used than just to check if it's placed somewhere. Worse, unless my logic is seriously flawed, some of the use all the pieces on a given level cards can only be gotten by putting one or more leftover pieces that cannot be used in some random corner. Otherwise, though, it's good. I wouldn't even have noticed that one problem if I weren't such a completionist.
If you go into the level editor and manually edit the string so that there's at least one test with an accept/reject output and at least one test with a specific string output, it will darken the screen and superimpose over it in the upper-left-hand corner "flixel v2.23 (release)" and below that a few messages about invalid tests. In the upper-right, it will show the current fps value. There isn't any apparent way to return the screen to normal short of refreshing the page, but all the buttons still work. However, it does prevent you from typing in the level text; apparently it learns.
I wonder how the malevolence engine works? I mean, I can see how one could parse a solution into code, but then how do you check if an algorithm works properly? I can't think of how to even start analyzing an algorithm, even using a language like Lisp where functions are themselves data.