Neuralink BCI After 28 Months: Arbaugh Tells Robotics Summit What Published Research Cannot

Noland Arbaugh — the first human being to receive a Neuralink brain-computer interface implant — closed the 2026 Robotics Summit & Expo in Boston on Thursday with a live demonstration that most of the engineers in the room had never seen outside a published paper: a man moving physical chess pieces using only his thoughts. The session, held at 3:30 p.m. ET at the Thomas M. Menino Convention and Exhibition Center, drew the summit's final standing ovation — from an audience of robotics developers who had spent two days asking how to get machines to respond the way human bodies do.

Twenty-eight months after his surgery at Barrow Neurological Institute in Phoenix, Arbaugh brought something to that audience that no clinical trial report can supply: a practitioner's account of what the technology actually does over time, how it fails, how it recovers, and what it feels like to be the human in the feedback loop.

What Noland Arbaugh Demonstrated at Robotics Summit 2026

Arbaugh's closing keynote, titled "Rewiring What's Possible: A New Era of Human Potential," was structured as a conversation with Steve Crowe, executive editor of The Robot Report. The centerpiece was a live demonstration of neural control applied to a physical chessboard — a system in which Arbaugh's motor-cortex intentions, decoded by the N1 chip and transmitted wirelessly via Bluetooth, directed the movement of physical chess pieces in real time.

The demonstration is not new for Arbaugh. He has been playing chess this way for more than a year. What made it significant for the summit's audience was the context: the engineers who watched it are building robots they hope will eventually receive similar motor-command streams. Arbaugh's implant, running for 28 months in a human body under daily conditions, is the closest thing to a field test of that proposition.

How Neuralink Brain-Computer Interface Decodes Movement

The Neuralink N1 implant that Arbaugh received in January 2024 uses 64 ultra-thin electrode threads, totaling 1,024 electrodes, implanted into the motor cortex — the brain region responsible for planning voluntary movement. When Arbaugh thinks about moving his hand, neurons in that region fire in patterns the implant captures, digitizes, and transmits to an external device via Bluetooth Low Energy. A machine-learning algorithm trained on his individual neural patterns then translates those signals into continuous cursor movement or, in Thursday's demonstration, physical chess-piece control.

The system's key distinction from earlier BCI approaches is resolution and bandwidth. The motor cortex does not produce a single on-off signal; it produces a continuous spatial-temporal pattern across hundreds of neurons that collectively encode direction, velocity, and intent. Kip Ludwig, a neural engineering expert, assessed the technology upon Arbaugh's first public demonstration in March 2024 as promising but still at an early stage — a calibration that 28 months of continuous use has substantially updated.

Arbaugh does not move. No physical action is involved. The signal path runs: motor-cortex intention → electrode capture → on-chip digitization → Bluetooth transmission → decoding algorithm → output command. For the robotics engineers in the room Thursday, that signal path is the part of the system they care most about: the precise point where neural intent becomes machine instruction, without any physical interface layer.

Thread Retraction: The Engineering Setback That Became a Data Point

About a month after surgery, approximately 85% of Arbaugh's electrode threads retracted from brain tissue — the wires that had been inserted with precision by the R1 surgical robot pulled back from their targets as the brain's immune response treated them as foreign objects. The incident dramatically reduced the number of active recording channels and impaired the system's performance.

"It sucked," Arbaugh said publicly. "It was really, really hard."

What happened next is the data point that distinguishes a 28-month trial from a three-month one. Neuralink's engineers resolved the retraction without additional surgery — by refining the software decoding algorithm to extract more useful signal from the surviving electrode population. Arbaugh's performance, by his own account, eventually exceeded his pre-retraction benchmarks. No subsequent patients in the PRIME clinical trial experienced retraction at the same scale, and Neuralink has since changed its surgical technique for later implants.

For robotics developers designing systems that receive continuous neural command streams, the thread-retraction episode illustrates an engineering reality that does not appear in early-stage trial data: the interface degrades under biological conditions, and the path to reliability runs through software adaptation, not only hardware precision.

10 Hours Daily: What Sustained BCI Use Means for Robotics Design

By August 2025, Fortune reported that Arbaugh was using his implant approximately 10 hours a day — for chess, for university coursework in neuroscience at a community college in Arizona, for scheduling, for communication. He named the implant Eve and describes it as his "brain co-pilot."

For robotics engineers, 10 hours of daily active use under real-world conditions is meaningful in a way that laboratory sessions are not. It captures the kind of variability — neural signal drift, user fatigue, adaptation over time, interference from environmental factors — that short lab protocols cannot reproduce. It establishes that a human user can sustain a continuous neural-control interface across tasks that vary in cognitive demand and motor-command complexity without encountering a fundamental usability ceiling.

Arbaugh's testimony at the summit carried a practical implication: the adaptation curve for a neural interface is long, non-linear, and ultimately recoverable from setbacks that would appear catastrophic in the early data. That is not a claim any published paper from the first year of the PRIME trial could support, because no one had yet lived inside the system long enough to know.

Beyond Arbaugh: Neuralink Trial Expands to More Than 20 Patients Worldwide

Arbaugh was Patient 1 in Neuralink's PRIME Study — Precise Robotically Implanted Brain-Computer Interface — an ongoing clinical trial targeting adults with quadriplegia from spinal cord injury or ALS. As of early 2026, the trial had enrolled more than 20 participants across sites in the US, UK, Canada, and United Arab Emirates. A subsequent UK patient reportedly controlled a computer within hours of surgery using the updated N1 architecture, which now uses 128 thinner threads with 8 electrodes each — maintaining 1,024 total electrodes while reducing the tissue displacement that contributed to the original retraction in Arbaugh's case.

Competitor Synchron's Stentrode takes a fundamentally different approach: implanted endovascularly through the jugular vein without opening the skull, it avoids cranial surgery entirely at the cost of electrode density. Precision Neuroscience, co-founded by a Neuralink alumnus, pursues a flexible surface array that sits on the cortex rather than penetrating it. Neither approach yet matches Neuralink's electrode count, and by extension, the resolution available for decoding complex motor intentions.

On December 31, 2025, Elon Musk announced that Neuralink would move to high-volume production of the N1 implant and a nearly fully automated surgical procedure. In May 2026, the company announced that its next-generation R1 robot can now place electrode threads into virtually any brain region — expanding potential clinical targets beyond motor-function restoration to include Parkinson's disease, refractory epilepsy, and treatment-resistant depression.

The BCI sector Arbaugh represents received a regulatory marker in May 2025, when the FDA awarded Neuralink Breakthrough Device Designation for a speech-restoration application covering patients with ALS, stroke, spinal cord injury, cerebral palsy, and multiple sclerosis. Phase 3 trials are expected to begin later in 2026, with a premarket approval submission targeting 2027 and commercial availability for paralysis patients projected for 2028.

What Does Neuralink Mean for Robotics Development?

The conversation that Arbaugh's presence opened at the 2026 Robotics Summit was not primarily about assistive technology. It was about architecture. Humanoid robots on the summit floor — including Boston Dynamics' electric Atlas, now in commercial shipment to Hyundai and Google DeepMind, and Agility Robotics' Digit, which has moved more than 100,000 totes at a GXO Logistics facility in Georgia under a Robots-as-a-Service model — receive their commands through programmed interfaces or trained AI models. A BCI-driven control layer would change that architecture at its root: the human operator's motor cortex would issue commands directly to the robot's control system, with the intention-to-action latency reduced to the time it takes a neural signal to cross a Bluetooth channel.

That is still a research proposition, not an available product. But Arbaugh's demonstration made it visible in a specific, non-hypothetical way. When the neural control of a physical object — a chess piece, a cursor, an exoskeleton joint — runs through a decoding algorithm in real time, the engineering question shifts from "can this work?" to "how do we make it reliable enough to deploy?" That is exactly the question the rest of the summit was asking about humanoid robots.

The Robotics Summit & Expo, produced by The Robot Report and WTWH Media with strategic partner MassRobotics, drew more than 5,000 engineers and developers across two days and 50-plus sessions. Arbaugh's keynote was its last event. The question his closing session posed — where is the boundary between operating a robot and being one — is not one any single summit will settle.

Frequently Asked Questions

How does the Neuralink brain-computer interface work?

The Neuralink N1 implant places 1,024 electrodes across thin flexible threads into the motor cortex, the brain region that plans voluntary movement. When a user thinks about moving, those neurons fire in patterns the electrodes capture and transmit wirelessly via Bluetooth to an external decoding algorithm, which translates the neural signals into cursor movement, keyboard input, or device control — without any physical action from the user.

What can Noland Arbaugh do with the Neuralink implant after 28 months?

As of mid-2025, Arbaugh was using his implant approximately 10 hours a day for activities including playing chess with a physical chessboard, studying neuroscience coursework, managing scheduling, and communicating — all controlled by thought alone. He also recovered from a significant early setback in which roughly 85% of his electrode threads retracted from brain tissue, an issue Neuralink resolved through software rather than additional surgery.

What is Neuralink's PRIME clinical trial?

The PRIME Study — Precise Robotically Implanted Brain-Computer Interface — is Neuralink's first-in-human clinical trial, targeting adults with quadriplegia from spinal cord injury or ALS. As of early 2026, it had enrolled more than 20 participants across sites in the United States, United Kingdom, Canada, and United Arab Emirates.

How does brain-computer interface technology connect to robotics development?

Researchers and engineers see neural implants as a potential direct-control layer for robotic systems: instead of programming a robot or training an AI model to interpret user commands, the operator's motor-cortex signals would issue commands directly to the robot's control system. This remains a research proposition, but Arbaugh's demonstrated ability to drive physical objects through continuous neural intent gives the robotics development community a concrete, operational data point to work from.