I am not sure you got my point. I'll try again. To efficiently search a space, you need some variation in the trials that are performed. However, that variation does not necessarily need to be embodied in the genomes of intelligent agents. It could be in the form of variations in lab experiments performed. Progress today does not depend on genetic variation between humans. It depends on memetic variation - and the memes are usually not embodied as agents that are conscious or do much cooperating. As far as I can tell, if you understand this, your original questions seem to fall apart.