The gap between biology and hardware
Biology signals with ions and molecules through soft, dynamic, self-built matter, in which hardware and software are entangled and barely understood. Silicon operates with electrons through rigid inorganic matter: scalable, resilient, and fixed in architecture once it leaves the fab. Something has to cross between them, and today, invasive or not, the crossing is far from ideal.
Microelectrode arrays read a blur of many neurons at once, and the tissue reacts to them. Optogenetics and sonogenetics are genetically encoded, but even at their best they resolve a single neuron, still far coarser than the computation going on inside one. Against what a cortex actually does, the resolution on offer, even with wishful thinking, is a 16-pixel screen against an 8K film, and the bandwidth is as lopsided as pre-fiber cable internet, fast in one direction and a trickle in the other.
Through my career I have been building the foundational stack for biology's upper-limit BCIs: interfaces that operate at the resolution of a single molecule, a single synapse, the way neurons interact with neurons, and that borrow the properties of both biology and modern hardware. Protein materials, so the interface can be built bottom up by cells and manufactured top down, to transduce an ionic signal into an electron event, cross the membrane, conduct over long range, and persist under biological and synthetic conditions alike. Electrogenetics, so a cell grows its own contact instead of having one implanted. And persistent biologics, so what is built holds up for years, in a body and at the interface and scale of modern hardware. Each field I worked in added one piece.