I love this. I'd like to see this structure applied to more things.

I also wonder if it can be applied to things where the system doesn't know the answer; an answer that's hard to generate but easy to verify. That probably only makes sense if participants can't fully verify it themselves though.