AI-Driven Synthesis-First Paradigm to Close Materials Discovery Gap

A new paper on arXiv (2605.00313) argues that AI-driven materials discovery is stalled by a synthesizability gap, where thousands of candidate structures cannot be made. The authors propose a synthesis-first paradigm treating executable protocols as primary design variables. Their roadmap has three pillars: representing synthesis procedures as machine-readable protocols, using generative and inverse-design models to propose reaction pathways, and integrating closed-loop optimization for experimental and sustainability constraints. The causal framework P->X->y links protocol P to structure X and properties y. The paper outlines methodological building blocks, standards, and self-driving laboratories.