The article discusses a study on predicting software faults using just two metrics: lines of code and cyclomatic complexity. The researchers used radial basis functions with Gaussian kernels as classifiers. They analyzed results from two NASA systems and found that these parsimonious classifiers can effectively assess software quality.