I mean, more around the reluctance to publish any more code publicly since it just gets sucked up by companies to train their models.