A mutation is the application of an operator against a subject, followed by running the selected tests. If the tests stay green the mutation is alive (bad — unspecified semantics, an automatically-derived flag equivalent to a reviewer asking why this change is safe). If tests fail the mutation is dead (good — semantics are nailed). Alive mutations are the primary artifact practitioners work with; each represents a latent regression a future coworker, intern, or your future self could introduce unnoticed.