Cameron R. Wolfe, Netflix staff scientist, compiles research showing world-model co-training improves agent planning and generalization · Digg