Representation is not a goal in and of itself.  To be useful, a knowledge base must allow us to get answers to interesting questions, in an effective way.  Unfortunately, this is not so easy.  The complexity of reasoning tasks has been a long-standing problem in AI.  It was given a physical manifestation by Henry Kautz in his 1989 Computers and Thought talk, and reincarnated by Stuart Russell 6 years later.  Now, six years after that, I’d like to reintroduce the complexity monster.