As interesting as this article was -- especially as he got into this from studying the neuroscience of bird song, something I was involved with years ago -- I think it's a stretch to call this 'engineering'.
It is an interesting take on an old technique. Instead of using direct electrical stimulation to stimulate the brain, he uses virally-transcoded neurons to respond to different wavelengths of light....then pipes a fiber optic cable into a mouse brain. To do what? To make it run in circles.
It's a proof of technology, but nothing more. Engineering the brain would imply we understand how it works, which, more or less, we still don't. Not really at a cellular level, not really at a systems level, not even really at a gross level either. We know an order of magnitude more than we did even a decade ago, but we are no closer to altering behavior than we were when the lobotomy was invented...the first 'neuroengineering'.
I think it is much more likely that we will first have engineered modules, either synthetic neuronal or otherwise, that will process independently and then 'plug into' our pre-existing sensory input pathways, rather than direct brain modification.