Why go inside
Most brain-computer interfaces that you can wear, like an EEG cap, listen through the skull. That is a bit like standing outside a stadium and guessing the score from the roar of the crowd. You can tell something big happened, but you cannot hear any single voice. Intracortical electrodes do the opposite: they sit *inside* the brain tissue, close enough to overhear individual neurons.
Why is that worth brain surgery? Because hearing single neurons fire gives the richest, fastest control signal we know how to collect. From the firing patterns of a few hundred cells in motor cortex, a decoder can reconstruct the direction and speed a person is *trying* to move a hand, even when the hand cannot move at all. That fidelity is the payoff that justifies an implant.
The Utah array
The classic implant is the Utah array, a microelectrode array developed in the early 1990s. Picture a tiny bed of nails about the size of a baby's fingernail: roughly one hundred stiff silicon needles, each around a millimeter long, packed onto a square base. Surgeons press it into the motor cortex so the tips end up among the neurons.
Each needle tip records the tiny voltage blips of nearby cells. A processing step called spike sorting then separates those blips into the firing trains of individual neurons. For about two decades this has been the workhorse of human implant research: when you read that someone moved a cursor or a robotic arm with their thoughts, a Utah array was very often the device doing the listening.
Flexible threads & high channel counts
A rigid array has two stubborn drawbacks. First, the brain is soft and constantly pulses with your heartbeat, while silicon is stiff, so the tissue chafes against an unyielding object. Second, a hundred channels is a narrow keyhole onto a brain with tens of billions of neurons. The newer wave of implants tries to fix both at once.
The general trend has three threads, so to speak. Make the electrodes thin and flexible so they bend with the tissue instead of fighting it. Push channel counts up from about a hundred toward the thousands, listening to more neurons at once. And go wireless, so signals leave the head without a connector poking through the skin, which is a major route for infection. Several companies and academic groups are pursuing devices with thousands of channels along these lines.
Biocompatibility & longevity
Here is the hard, unglamorous problem that no implant has fully solved: the brain treats the electrode as a splinter. This is the foreign-body response. Immune cells called glia migrate to the device and gradually wrap it in a dense sheath of scar tissue, walling it off from the very neurons it is meant to hear.
The consequence shows up slowly. Over months and years, the crisp spikes of individual neurons get fainter and harder to separate, until the array hears mostly a muffled hum instead of distinct voices. A device that gave beautiful single-neuron control on day one may, channel by channel, fade toward silence. Flexible threads are partly an attempt to provoke a gentler scar, but no one has yet demonstrated a recording that stays sharp and stable for decades.
This is why durable implants remain an open frontier rather than a finished product. A truly useful clinical device must last a lifetime, survive immune attack, never leak or infect, and keep decoding as signals drift. Progress is real, but honesty matters more than hype here: the engineering of a brain implant you could trust for thirty years is still unsolved. That gap is exactly what stands between today's research devices and routine clinical use.