This talk delves deeper into the topics discussed at 2009's [[MindTheSemanticGap|Mind the Semantic Gap] workshop. As the performance of our systems improves, the performance model tends to degrade. In other words, our ability to predict, or even measure, the performance of a particular piece of source code tends to get worse. What can we do about this, as programmers, performance analysts, language designers, and VM designers, and Computer Science Educators? Come ready to share your opinions and war stories.