Many questions in molecular science and engineering rely centrally on the problem of molecular solvation: how strongly a solute molecule (for example, a protein) interacts with the surrounding liquid (solvent). Continuum electrostatic models based on macroscopic dielectric theory and the Poisson--Boltzmann equation provide surprisingly accurate insights, and fast numerical algorithms enable calculations in minutes, but unfortunately the model's approximations limit its value in many applications. On the other hand, fully atomistic simulations with explicit solvent molecules can give quantitatively accurate predictions, but require orders of magnitude more computational resources (thousands to millions of CPU hours). Recognizing that the main errors in the continuum model arise from the "solvation layer" -- the first layer of solvent molecules around the protein, which do not behave like those far away -- we introduced a nonlinear correction to the traditional macroscopic interface condition. Our new solvation-layer interface condition (SLIC) model provides quantitative accuracy for numerous problems in which the traditional continuum model fails totally: examples include temperature dependence, pH dependence, solvent mixtures of varying composition, and charge-sign asymmetry. After describing the SLIC multiscale boundary condition and its improved performance, we will highlight opportunities for both applications and further theoretical development.