An MIT AI framework analyzed scientific literature and over one million rock samples, sorting candidate cement replacements into 19 material types. The most promising replacement: old ceramic waste — tiles, bricks, pottery.
Crushed ceramics have high pozzolanic reactivity — they react with calcium hydroxide in cement to form additional binding compounds. This property was observed and exploited in ancient Roman concrete, which used crushed ceramics for waterproofing. Roman harbor structures in Caesarea and Baiae have survived two thousand years of seawater immersion.
The AI, trained on modern data and optimizing for material reactivity, convergence stability, and global availability, independently converged on a technique the Romans discovered empirically. The “new” sustainable concrete is the old concrete.
Suitable replacement materials are globally available. Ceramic waste from construction demolition, tile manufacturing rejects, and obsolete brickwork requires only grinding before use — no chemical processing, no kiln firing, no new energy input.
The convergence is meaningful. When an optimization algorithm with access to a million samples arrives at the same solution as artisans operating two millennia ago, it suggests the ancients were optimizing the same objective function with different tools. They lacked materials science but had centuries of empirical selection pressure. Both paths — gradient descent and cultural evolution — found the same basin. The optimum was always there; what changed was how we climb toward it.