I've started to think I'm better off building lots of small OSS libraries than one big one.
The big one seems like some massive accumulator bet, which is really unlikely to pay off, whereas one of the smaller parts may just have an impact on it's own...