From the extended thread, the idea here is that larger companies are less capable of experimenting, so it's good to experiment before scaling up.