A second, quieter revolution is unfolding at the intersection of AI and the life sciences. Following breakthroughs in protein structure prediction, a new wave of generative models is moving past predicting how proteins fold to designing entirely new ones: enzymes, antibodies, and binding molecules that do not exist in nature, built to specification in silico before ever touching a lab bench.
These generative biology tools compress a process that once took years of wet-lab trial and error into weeks of computational search. Pharmaceutical teams are now using AI-guided design to propose drug candidates, predict how a molecule will behave inside the body, and flag likely toxicity issues long before costly clinical trials begin, materially shortening the path from hypothesis to therapy.
The open questions are no longer purely computational. Validating AI-designed molecules still requires rigorous experimental confirmation, and regulators are actively working out how to evaluate medicines whose starting point was a model's prediction rather than a chemist's intuition. Even so, the direction is clear: the next generation of therapeutics is increasingly being sketched first in code.