I will give an overview of what we know (and don't know) about ("long code") tests separating dictatorships from functions far from dictatorships. Such tests play a role in proving hardness of approximation for constraint satisfaction problems.