Discovery of useful or interesting new materials and molecules by computation requires an efficient, accurate, and reliable theoretical method, and the preferred method is still Kohn-Sham density functional theory[1]. In this theory, the exact ground-state energy and electron density (and thus the nuclear positions) can be found by solving self-consistent one-electron equations. The exchange-correlation energy as a functional of the electron density must in practice be approximated. I will discuss a systematic and proven way to improve the approximations, making them more accurate and reliable at a modest increase of computational cost. Then I will show how this approach has led to SCAN [2], a strongly-constrained and appropriately normed functional that, without being fitted to any bonded system, makes accurate predictions for diversely-bonded materials and molecules [3]. [1] W. Kohn and L.J. Sham, Self-Consistent Equations Including Exchange and Correlation Effects, Phys. Rev. 140, A1133 (1965).[2] J. Sun, A. Ruzsinszky, and J.P. Perdew, Strongly Constrained and Appropriately Normed Semi-local Density Functional, Phys. Rev. Lett. 115, 036402 (2015).[3] J. Sun, et al., Accurate First-Principles Structures and Energies of Diversely-Bonded Systems from an Efficient Density Functional, Nat. Chem. 8, 831 (2016).