Creating an alloy from several metals is like cooking a soup from mysterious ingredients: guessing the final texture without trial and error is nearly impossible. Conventional models often get it wrong when there are more than three different kinds of atoms. The new method replaces endless experiments with computation: the more atomic arrangements tested, the higher the information entropy — a measure of how much of the ‘recipe’ space has been explored. A neural network trained on examples, like a seasoned taster, quickly evaluates millions of combinations, while an ‘alchemical’ trick lets you mentally swap one element for another right during the calculations.

When entropy stops growing, the search is called off: all reasonable options have been exhausted. This way, the structure can be predicted even for five-component mixtures like the Cantor alloy (aluminum-chromium-iron-nickel-cobalt). Even steel — just an alloy of iron and carbon — pales next to this atomic cocktail. Amazingly, the same math that helped compress a phone call into bits is now paving the way to super-strong materials for aviation.