The computation of next-to-leading order QCD corrections is important to new physics searches at the LHC. A key and difficult ingredient in any such correction is the one-loop amplitude for the relevant process. In this talk, I describe BlackHat, a C++ implementation of a primarily-numerical approach to computing one-loop amplitudes. The approach is based on the on-shell method for computing amplitudes, relying on unitarity, on-shell recursion relations, and the spinor-helicity method. I illustrate the approach by presenting the BlackHat Collaboration's recent results for W+3 jet distributions.