What would it take to engineer a brain circuit to perform a new kind of computation or to augment an existing brain computation with additional information? Perhaps you could augment a memory circuit (starting with, say, a mouse) so that it could tap into digital data, boost the capacity of working memory so that dozens of things could be held in mind at once, or enable algorithms from computer science to be run on in-brain wetware. A key difference between neural circuits and computers, of course, is that computers were designed by humans, so the principles of how to program them are well-defined. However, the principles of controlling neural circuits, to make them do exactly what you want them to do, are not fully understood. Perhaps you activate neurons in a certain pattern, and as-yet-unknown homeostatic mechanisms kick in and cancel out the effect you just created. Perhaps you drive one kind of signal at a synapse, and a chemical cascade is triggered that sends a novel signal in a retrograde fashion. We don’t have a full list of the cell types of any mammalian brain, so perhaps you perturb one kind of cell, and an as-yet-undescribed cell type, equipped with unknown mechanisms, rebels against the changes you were trying to induce. Right now, there are many ongoing attempts to make maps of the brain’s wiring, or connectomics. That is a necessary step toward understanding brain networks. But if we want the brain to become a predictably engineerable system, we will likely need to go further and map out the molecular and cellular mechanisms throughout the wiring.