It is a pleasure to help celebrate and remember Axel Becke, and particularly his pioneering contributions to the development of modern density functional theory. In this talk, I will try to do so by describing some recent developments aimed at improved density functionals from my group. None of this work would have been possible without Axel’s inventions. In particular, seeking the broadest accuracy that is possible for hybrid density functionals, I will briefly summarize our new Carefully Optimized and Appropriately Constrained Hybrid (COACH). I will then discuss new work that aims to provide DFT calculations with error estimates, and show that transferable and useful error estimates can be obtained using statistical theory applied to generalized basis functionals (i.e. modern hybrids). These error estimates can also be applied to energy differences, and even differences of energy differences, to reflect the extent of error cancellation.