Phase transitions in random matrix ensembles, notably those involving extreme eigenvalues, illuminate how macroscopic behaviour emerges from intricate microscopic correlations.