It seems fairly plausible that they will become useful. I would not consider coding requiring vastly better AI models than the best we have today. The more interesting question is what advantages excess intelligence provides in the future more generally (warfare, system management, etc)