Researchers have trained an end-to-end differentiable neural cellular automata model to simulate morphogenesis. This self-organising system is capable of both growing and regenerating specific target patterns. The development offers promising applications for creating highly adaptive and self-repairing digital structures.