Numerical relativity has been spectacularly successful during the past decades, with numerous contributions toward gravitational wave astrophysics for binary black holes and neutron stars. Despite all accomplishments, the increasing sensitivity of gravitational wave detectors and the broader bandwidth of future detectors require significant further improvements of the numerical codes, both for black holes and neutron stars. In this talk, I will use the vacuum case of binary black holes to illustrate several dimensions along which such improvements are being pursued within the SXS collaboration. These dimensions include gravitational wave extraction, new computational algorithms, longer simulations to cover broader detector bandwidth, as well as an extension of the parameter space coverage to eccentric binary black holes and hyperbolic encounters.