header-only, plugs into existing generators with little effort

for small decision trees, rapidly converge to the desired behavior

for huge ones, slow convergence is fine

support weighted choices

allow the user to customize behavior, e.g. don't insist that an integer constant in the output corresponds to 2^64 leaves

write a collection of trivial but pathological generators, such as the one that calls flip() 1000 times in succession

work out how to evaluate how well this thing works, when used in non-trivial situations

plug this into Csmith and YARPGen and see what happens

work out how to take hints from the user about things like desirable patterns, subtrees that are likely to behave similarly every time we reach them, etc.

work out how to take feedback about good/bad paths into account