Generative design

Generative design is a computational approach to engineering and architecture in which algorithms explore a vast space of possible designs and select those that satisfy specified constraints, rather than having a human designer specify a single solution. The method treats design as a search problem: the designer specifies what the artifact must do, and the algorithm discovers how it can be done. Generative design has produced structures — from aircraft components to building facades — that are lighter, stronger, and more complex than anything designed by human intuition alone.

The approach is particularly powerful in domains where the relationship between form and function is too complex for analytical solution, such as topology optimization and swarm behavior design. However, generative design does not eliminate the design gap; it displaces it. The creativity moves from the human designer to the fitness function, and the challenge becomes specifying what good means in a way that the algorithm can optimize. A poorly specified fitness function produces artifacts that satisfy the letter of the constraint while violating its spirit — a phenomenon known as specification gaming or reward hacking.