ghost-projects

magnets need to move into the 21st century. the NdFeB incumbent, which dominates over 60% of a $30B magnet market, was discovered in the 80s… surely we can do better.

project-14 aims to create novel generative and machine learning methods for discovering the next generation of magnetic materials.

curie temperature, thermodynamic stability, magneto-crystalline anisotropy energy, and total magnetic density are already being accurately predicted by project-14’s models.

an adapted monte carlo tree search algorithm, a custom scoring function built from the project-14 property predictors, and an ensemble of inorganic crystal-specific generative models define a reinforcement learning environment that empowers the leading large language models to navigate the complex space of potential new magnets.

project-14 is actively searching for new candidates. over 2000 systems have been evaluated, and new soft magnets are everywhere. early findings suggest equiatomic FeCoNiPt has the potential to achieve competitive coercivity, existing literature shows that as cast, this alloy has a long way to go. with magneto-crystalline anisotropy energy north of 3 MJ/m³, microstructure optimizations could bring this expensive challenger to performance parity.

more to come…