Next-generation continuum surveys will be strongly constrained by dynamic range and confusion. The ASKAP-EMU (Evolutionary Map of the Universe) team is exploring ways of using multiwavelength data to penetrate well beyond the classical confusion limit. For example, deep IR surveys (from Herschel, VISTA, etc) can be used to identify all sources with a star-forming SED, and then the radio-IR correlation can be used to subtract all these sources from the radio image. The remaining difference image contains only those sources (mainly AGNs) which do not have a star-forming SED. In this talk I will describe the EMU project (which proposes to image 75% of the sky to 10$\mu$Jy rms, and 30 sq deg to 1 $\mu$Jy rms), the likely constraints and issues which are now being explored by the EMU team, and then focus on confusion issues and how we might penetrate well below the classical confusion limit.