How brain-computer interfaces work: Neuralink and the future of neurotech

01 What a brain-computer interface actually does

A brain-computer interface (BCI) establishes a direct communication pathway between the brain's electrical activity and an external device — most commonly a computer cursor, a robotic arm, or a speech synthesizer. The core idea is straightforward: neurons communicate by firing tiny electrical impulses, and a BCI detects those impulses, decodes the patterns, and translates them into commands a machine can execute.

This is not telepathy or mind reading in the popular sense. The user does not broadcast abstract thoughts. Instead, they learn to produce specific, repeatable neural firing patterns that the decoding algorithm has been trained to recognize. Early demonstrations required patients to imagine moving their hands, which activates motor cortex neurons in predictable ways. The decoder maps that activity to a cursor or prosthetic limb.

The field emerged from decades of neuroprosthetics research. Cochlear implants, which restore hearing by directly stimulating the auditory nerve, are arguably the most successful BCI in history — over 900,000 have been implanted worldwide. Deep brain stimulation for Parkinson's disease, which uses implanted electrodes to deliver targeted electrical pulses, is another proven application. What distinguishes the current wave of BCI development is the ambition to create bidirectional, high-bandwidth links capable of far more than sensory substitution or tremor suppression.

02 How Neuralink's implant works

Neuralink, founded by Elon Musk in 2016, has developed a coin-sized implant called N1 that sits flush with the skull beneath the skin. The device contains 1,024 ultra-thin electrodes distributed across 64 threads, each thread approximately 5 micrometers wide — thinner than a human hair. These threads are inserted into the cortical surface by a surgical robot Neuralink calls R1, which is designed to place the threads with micron-level precision while avoiding blood vessels.

The threads record activity from individual neurons. Signals travel through the implant to a wireless transmitter, which sends data to a receiver worn behind the ear. Custom software decodes the neural activity in real time, translating intended movements into cursor control on a computer or mobile device. The patient interacts with the system through an app, calibrating the decoder by imagining specific movements until the mapping becomes intuitive.

Neuralink's technical advantage is electrode density. Where earlier systems used arrays of fewer than 100 electrodes, the N1's 1,024-channel capacity theoretically allows for higher-resolution decoding. The company has also emphasized hermetic sealing and wireless operation, eliminating the need for external cables or percutaneous connectors that have historically been infection risks in long-term implants.

03 The surgical procedure and risks

The implantation procedure begins with a craniectomy — a small section of skull is removed to make room for the device. The R1 robot then inserts the flexible polymer threads into the cortical tissue, typically targeting the hand knob region of the motor cortex for initial trials. The entire surgical process is designed to take less than an hour and patients are typically discharged the same day.

The risks are real. Brain surgery carries inherent dangers of hemorrhage, infection, and seizures. Beyond the surgical phase, long-term implantation introduces concerns about tissue scarring around the electrodes, which can degrade signal quality over time. There is also the risk of device migration, where the threads shift position within the brain, potentially causing new damage or rendering the system non-functional.

Neuralink's first human patient, Noland Arbaugh, received the implant in January 2024. He demonstrated the ability to control a computer cursor and play video games using only his thoughts. However, within weeks, many of the threads retracted from the brain tissue, reducing the number of functioning electrodes. Neuralink reportedly addressed this in a second patient through software improvements and modified surgical techniques, though the long-term stability of the threads remains an open question.

04 Current clinical trial results

Neuralink's PRIME study (Precise Robotically Implanted Brain-Computer Interface) received FDA approval for human trials in May 2023. The first implant was performed in January 2024, with a second patient in August 2024. Results have been promising in functional terms — patients have achieved cursor control sufficient for web browsing, messaging, and gaming — but the electrode degradation problem revealed in the first patient underscores how far the technology is from clinical maturity.

The performance metric that matters is bits per second (BPS), the rate at which neural signals are decoded into actionable commands. Neuralink's patients have reportedly achieved BPS scores in the range of 4-8 bits per second, which is competitive with or exceeds prior academic results from the BrainGate consortium. For context, able-bodied individuals using a mouse achieve roughly 10-15 BPS.

The competitive landscape includes Synchron, which has implanted its Stentrode device in over ten patients since 2019. Blackrock Neurotech has the longest track record, with over 40 implantations using its Utah Array technology, some functioning for over seven years. Each approach makes different trade-offs between invasiveness, signal quality, and durability.

05 Competing approaches from Synchron and others

Synchron takes a fundamentally different approach. Rather than opening the skull and placing electrodes directly on cortical tissue, Synchron's Stentrode is delivered through the vascular system. A catheter threads the device through the jugular vein to a blood vessel resting against the motor cortex. Once in position, the stent-like device expands and the electrodes record neural activity through the blood vessel wall.

This endovascular approach eliminates craniectomy and direct brain penetration, dramatically reducing surgical risk. The trade-off is signal quality: the Stentrode records from a distance, capturing lower-resolution signals than Neuralink's intracortical threads. Synchron's patients have achieved BPS scores of roughly 2-4 bits per second — enough for text selection and basic cursor control, but below Neuralink's peak performance.

Other companies pursue different niches. Paradromics is developing a high-data-rate implant aimed at restoring communication for patients with ALS, targeting the same Utah Array technology refined for higher channel counts. Precision Neuroscience, founded by former Neuralink executives, is developing a flexible film array that sits on the cortical surface without penetrating tissue — a less invasive option that may reduce scarring risks. Blackrock Neurotech continues to iterate on its Utah Array, the gold standard in academic BCI research for two decades.

06 What BCI means for paralysis patients

For individuals with severe paralysis from spinal cord injury, ALS, or stroke, a BCI can restore a degree of independence that no existing technology provides. The primary application today is communication: patients who cannot speak or move their limbs can use a BCI to type, send messages, and browse the web. For someone locked in by ALS, the ability to select letters at even one per second is transformative.

Motor control is the second frontier. Neuralink's patients have demonstrated cursor control precise enough to play chess and Civilization VI. Researchers at BrainGate have decoded attempted handwriting movements, achieving typing speeds of 90 characters per minute — roughly double the speed of able-bodied smartphone typing. The ultimate goal is direct control of prosthetic limbs, though this requires far more complex decoding and, for bidirectional feedback, the ability to send signals back into the brain.

The patient population for these early trials is small and carefully selected. All current participants have had their conditions for months or years, ensuring stable neural profiles. Acute injuries present different challenges — the brain is in flux, and timing the implantation for optimal signal acquisition remains an open clinical question.

07 The ethical and privacy concerns

BCI technology raises privacy questions that conventional medical devices do not. A cochlear implant does not record your thoughts; a Neuralink thread potentially does. Neural data is intimate in ways that heart rate or blood pressure are not — it encodes intentions, emotional states, and cognitive patterns. Who owns that data? Can it be subpoenaed? Can insurers use it to assess mental health risk?

The question of autonomy is equally thorny. If a BCI translates neural activity into actions, who is responsible when the translation is wrong? A misdecoded command could send an email to the wrong person or move a wheelchair into traffic. The legal framework for assigning liability when a machine acts on behalf of a brain signal is essentially nonexistent.

Elon Musk has spoken about BCI as eventually enabling human-AI symbiosis, allowing humans to keep pace with artificial intelligence by effectively merging with it. This framing, while speculative, raises the specter of unequal access. If BCI becomes a cognitive enhancement tool rather than a medical device, the gap between those who can afford neural augmentation and those who cannot could dwarf existing inequalities in education and access to technology.

The governance gap is real. No federal agency has a clear mandate to regulate the long-term implications of consumer brain implants. The FDA approves medical devices for safety and efficacy, but it is not designed to address questions of cognitive privacy, algorithmic bias in neural decoding, or the societal effects of widespread neural augmentation.