Picture a person with a spinal cord injury who has not moved a hand in years. A surgeon makes a small opening in the skull, and a thin device is set against the surface of the brain. Later, the person thinks about moving a cursor, and the cursor moves. Nothing about that sequence is fantasy. It has been shown in research for more than a decade. What is new is the attempt to make it a routine procedure, performed by a robot, in a company's trial. That is the question this essay follows: what it takes to read and write signals in the brain, and what a trial registry can and cannot tell us about the answer.

What a brain-computer interface does

A brain-computer interface, or BCI, is a system that reads neural activity and turns it into commands for a computer or a machine. Most current research devices record the electrical activity of neurons with tiny electrodes. Software learns which patterns of activity match an intended movement, and then uses those patterns to control a cursor or a robotic arm.

The distinction that matters is between devices that sit outside the skull and devices placed inside it. Inside the skull, signals are clearer, but the surgery is more serious, and every electrode is a foreign object that the body can react to over years.

The earlier record

The idea is not new, and the published record is the best guide to what has already been shown. In 2011, Simeral and colleagues reported in the Journal of Neural Engineering on point-and-click control by a participant with tetraplegia, using electrodes placed in the motor cortex. Our reading of a secondary summary, not the full paper: the participant was tested about 1,000 days after implant, and the study reported a mean of 91.3 per cent correct target acquisition over five days. The primary text has not been checked for this essay.

A year later, Hochberg and colleagues published in Nature a study in which people with tetraplegia used a neurally controlled robotic arm to reach and grasp objects. The paper is widely cited as one of the first demonstrations that people with severe paralysis could use brain signals to control a prosthetic arm to perform three-dimensional reach and grasp.

Our reading: these studies show that the basic function works in small numbers of people. They do not show how long it lasts in most people, what it costs, or how it performs outside a research lab.

The registry entry

On 10 November 2026, a study was first posted to ClinicalTrials.gov under the identifier NCT07929792. It is titled "An Early Feasibility Study of a Robotically Implanted Brain-Computer Interface for Neural Control of External Devices in Spinal Cord Injury, Amyotrophic Lateral Sclerosis, and Chronic Ischemic Stroke," and its acronym is PRIME2. The sponsor is Neuralink Corp.

The registry lists an estimated enrolment of 10 participants. It names one site, the University of Miami, and its brief summary says the study may run at up to five sites in the United States. The device is the N1 implant, placed with an R1 robot. The conditions listed are spinal cord injury, amyotrophic lateral sclerosis, chronic ischaemic stroke and tetraplegia or tetraparesis. The study is in the recruiting phase, with an estimated start in November 2026. The registry record has no posted results.

These are the registry's own facts. They describe a plan. They say nothing about whether the device works, how safe it is, or what the participants will be able to do.

Why the robot matters

The surgical step is the part most likely to decide whether this becomes routine. A procedure that needs a skilled neurosurgeon for every case will be slow and expensive. A robot that places electrodes with a fixed, repeatable method could change the economics, but only if it is safe and reliable across many patients. The registry does not describe the robot's performance, and we have not seen any data on it.

Implications

Three groups have the most at stake.

The first is people with paralysis or ALS, who would gain the most if the technology works and who would bear most of the surgical risk. Early feasibility studies are designed to find out whether a device is safe enough to test further, not whether it helps patients.

The second is the people who would be asked to trust the data. A registry entry, a company announcement and a peer-reviewed paper are different kinds of evidence, and the public often receives them as if they were one.

The third is everyone else, once the technology exists. Speculation, labelled as ours: if reading and writing the brain become practical, the questions will move from who can afford surgery to who can refuse the data.

What could go wrong is familiar from other fields: a device that works in ten patients is sold to millions before anyone learns what happens in the tenth year.

Conclusion: the interface is the easy part

Reading a cursor from a brain is a demonstration. Living with an implant for years, in a body that changes, under a company that might be sold or fail, is the test. The registry entry tells us that someone has decided to try. It does not tell us whether the trial will show that the attempt was worth making. That answer will come from patients, not from press releases.

Sources: ClinicalTrials.gov, NCT07929792 (PRIME2; first posted 10 November 2026); Simeral et al., point-and-click cursor control with an intracortical BCI, Journal of Neural Engineering (2011), PubMed 21436513 (primary text not checked); Hochberg et al., "Reach and grasp by people with tetraplegia using a neurally controlled robotic arm," Nature 485, 372–375 (2012), doi:10.1038/nature11076.