It seemed to have a semi decent crawler and p2p system but it was beyond useless for searching. It was like using grep to search the internet. I just searched "youtube" and the first 10 pages would be random blogs that mention the word youtube instead of the actual website.
For me, the big problem of yacy is Java. I think that maybe the team of project work to implement the "algorith" in others languages (such as torrent protocol, that it is in C++, Java, Python...).