Despite rapid progress in the computational design of novel functional materials, the materials discovery pipeline remains bottlenecked by our inability to reliably synthesize predicted compounds in the lab. Developing a theoretical foundation for predictive materials synthesis requires a more quantitative understanding of metastable phases, which often appear as kinetic byproducts during materials formation. By mapping the thermodynamic landscape of crystalline metastability , and calculating relative nucleation rates between competing polymorphs , we can construct synthesis maps to navigate through the thermodynamic and kinetic energy landscape towards desired material phases. I will showcase several applications of this ab initio framework to predict non-equilibrium crystallization pathways of carbonate minerals and functional manganese oxides in hydrothermal synthesis, and conclude with thermodynamic strategies for the discovery and synthesis of metastable thin-film nitride semiconductors. Mastering metastability will deepen our fundamental understanding of nucleation and crystal growth, and can expand the search space for functional technological materials beyond equilibrium phases and compositions.