For years, brain-computer interfaces (BCIs) have sounded like something borrowed from science fiction: tiny electrodes reading electrical activity inside the brain and translating those signals into commands for a computer, a robotic arm or another device.

But the most important challenge facing this technology may not be the artificial intelligence decoding the brain signals.

It may be the surgery.

Musumeci Online – The Podcast. It is perfect for driving, commuting, or waiting in line!

In May 2026, Neuralink performed a significant step in that direction at Toronto Western Hospital in Canada. During a clinical trial, the company inserted its electrode threads through the dura mater without cutting or removing it. Neuralink describes this as its first transdural procedure. The participant was reportedly able to control a computer cursor using their thoughts within an hour of surgery.

At first glance, this may sound like a relatively small modification to a surgical technique.

It is not.

If the approach proves safe and reliable in larger numbers of patients, it could help address one of the biggest obstacles to the future of implantable BCIs: how do we move from impressive experimental demonstrations to a medical technology that can be implanted repeatedly, safely and at scale?

And this is where the story of Neuralink becomes much bigger than Elon Musk.

First, what exactly is Neuralink trying to build?

A brain-computer interface is essentially a communication system between the brain and an external device.

Our brains communicate using electrical activity. Every time we decide to move a hand, speak a word or perform another action, populations of neurons generate patterns of electrical signals.

Normally, these signals travel through the nervous system to muscles.

If a spinal cord injury or neurological disease interrupts that pathway, the brain may still generate the intention to move even though the muscles can no longer receive the command.

A BCI attempts to create a new route.

Instead of:

Brain → spinal cord → muscles

the system can create something more like:

Brain → electrodes → computer → external device

The external device could be a computer cursor, communication system, robotic arm or potentially, in the future, a much broader range of assistive technologies.

Neuralink’s N1 implant is designed around this principle. The current system uses 1,024 electrodes distributed across 64 flexible threads, each thinner than a human hair. A surgical robot places these threads into the brain’s motor regions.

The goal is not to “read someone’s mind” in the science-fiction sense.

It is much more specific.

The system tries to detect patterns associated with an intended action — for example, the intention to move a cursor — and translate those patterns into a command.

That distinction is important.

A BCI is not necessarily reading what you are thinking.

It is trying to decode particular patterns of neural activity.

The brain has armour — and it is called the dura mater

To understand why Neuralink’s latest procedure matters, we need to take a short trip through brain anatomy.

The brain is not sitting directly underneath the skull.

It is protected by three layers of tissue called the meninges.

From the outside inward, they are:

• the dura mater
• the arachnoid mater
• the pia mater

The dura mater is the outermost layer.

It is thick, strong and relatively tough — almost like a protective biological sheet covering the brain.

And that is precisely why it creates a problem for brain-computer interfaces.

Traditional invasive cortical implants generally require surgeons to open the skull and then access the surface of the brain. In procedures involving subdural or intracortical electrodes, the dura is opened or removed in the region where the electrodes will be positioned.

Neuralink’s earlier human procedures followed this general principle.

The new approach attempts something different:

leave the dura intact and pass the electrode threads through it.

In other words, instead of removing part of the brain’s protective layer to reach the cortex, the electrode threads pass through it.

It is a deceptively simple idea.

Making it work is anything but simple.

Why is going through the dura so difficult?

Imagine trying to insert a needle into a moving target while a protective curtain is covering it — and you are not allowed to lift the curtain.

That is roughly the engineering challenge.

The electrode threads used by Neuralink are extremely thin. According to the company, the dura can be more than 10 times thicker than the threads themselves. At the same time, the brain beneath the dura is not stationary: it moves subtly with breathing and the heartbeat.

There is another problem.

The dura and the tissues around it can obscure the small blood vessels on the brain’s surface.

For a surgeon, this creates an obvious challenge: an electrode must reach its target while avoiding structures that could cause bleeding or tissue damage.

Neuralink therefore had to solve several problems simultaneously.

It developed new insertion techniques, synthetic dura models for testing and imaging approaches designed to help visualise what is happening beneath the intact dura.

Among the technologies used are indocyanine green (ICG) video angiography, which can help visualise blood flow, and optical coherence tomography (OCT), which can provide precise measurements of the distance to the brain surface.

The important point is that this is not simply a new needle.

It is an entire surgical system involving robotics, imaging, neuroscience and precision engineering.

Why does preserving the dura matter?

The dura is there for a reason.

Leaving it intact could potentially reduce the amount of tissue manipulation required during implantation and eliminate one of the delicate surgical steps involved in traditional access to the cortex.

Neuralink describes the idea rather elegantly:

“The best step is no step.”

Removing the need for a durectomy could make the procedure more standardised and potentially easier to automate.

This matters because medicine has a long history of becoming safer when procedures become simpler and more reproducible.

Think about laparoscopic surgery.

Instead of making a large abdominal incision, surgeons learned to operate through small access points using cameras and specialised instruments. The technology did not simply make surgery more sophisticated; it changed the physical pathway through which surgeons reached the target.

Neuralink is attempting something conceptually similar.

The target remains the same.

The pathway is changing.

But there is a trade-off: signal quality

There is an important catch.

The closer an electrode gets to individual neurons, the richer and more precise the electrical information can potentially become.

The dura, however, is an additional layer between the electrode and the brain.

That means it can attenuate neural signals.

This creates a fundamental engineering trade-off:

more direct access can provide stronger signals, but greater invasiveness can increase surgical complexity and risk.

Scientific research comparing epidural and subdural electrocorticography has found that signals recorded above the dura can have lower amplitude, while the difference in practical signal decoding may be smaller than might initially be expected. A review published in the European Journal of Neuroscience concluded that both approaches could potentially be suitable for long-term recordings, depending on the application.

Earlier work also suggested that the dura does not necessarily prevent useful neural features from being detected, although subdural recordings can be more robust under certain noise conditions.

This is why Neuralink’s approach is particularly interesting.

The company is trying to obtain the advantages of intracortical recording — electrodes actually entering the cortex — while avoiding at least one part of the conventional surgical access process.

The question is whether this compromise can deliver the same level of useful neural information over the long term.

That is something clinical research still needs to establish.

Neuralink is not the only path to a brain-computer interface

One of the most fascinating aspects of today’s BCI field is that there is no single solution.

Different research groups are effectively exploring different points on the spectrum between signal quality and invasiveness.

Intracortical interfaces

Technologies such as the Utah Array have demonstrated for decades that tiny penetrating electrodes can record neural activity with high spatial resolution.

The Utah Array was first implanted in humans in 2004 and has been used in numerous research programmes involving computer control, communication and neuroprosthetics.

Neuralink’s system belongs broadly to this intracortical family, but its flexible threads and robotic implantation system represent a different technological approach.

The advantage is powerful neural information.

The disadvantage is that the electrodes must enter the brain.

Surface-based interfaces

Other systems record signals from the surface of the brain rather than penetrating deep into the cortex.

This can reduce tissue penetration, but the electrical signals may be less detailed.

Endovascular interfaces

Then there is another fascinating approach.

Synchron’s Stentrode, for example, is designed to reach the brain’s blood vessels through a catheter introduced via the jugular vein rather than through an open-brain surgical procedure. Its clinical studies are investigating whether this approach can restore control of digital devices in people with severe motor impairment.

Instead of asking:

“How can we put better electrodes into the brain?”

the endovascular approach asks:

“Can we reach the brain through the vascular system?”

And there is a fourth possibility.

Non-invasive BCIs

Companies and research groups are also investigating whether signals can be measured without implanting anything at all.

The advantage is obvious: no brain surgery.

The disadvantage is equally obvious: the skull acts as another barrier between the sensors and the brain, making the signals weaker and noisier.

Artificial intelligence can help compensate for some of this noise, but there is still a fundamental physical limitation.

This creates a fascinating race.

Some researchers are trying to improve the hardware.

Others are trying to improve the AI.

And others are trying to redesign the entire route between the brain and the machine.

Why Neuralink’s milestone could be more important than it looks

The most interesting part of Neuralink’s transdural procedure may therefore have little to do with the headline:

“Neuralink puts an implant into a brain.”

That has already happened.

The more interesting question is:

Can Neuralink make implantation easier to reproduce?

That is a completely different problem.

Imagine that a BCI works brilliantly but requires a highly specialised neurosurgical team, a complex procedure, extensive manual intervention and considerable operating-room time.

It could remain a remarkable scientific experiment without becoming a widely available medical treatment.

Now imagine that the procedure becomes increasingly standardised.

A robotic system performs more of the delicate steps.

Imaging helps the robot understand the anatomy.

Software guides the procedure.

The dura does not have to be removed.

The implant is wireless.

The operation becomes shorter and more predictable.

Suddenly, the conversation changes from:

“Can this technology work?”

to:

“How many people could we safely treat?”

That is the transition from laboratory innovation to healthcare technology.

From one extraordinary patient to thousands of ordinary clinical procedures

This distinction is particularly important in medicine.

A technology does not become transformative simply because it works once.

It becomes transformative when clinicians can use it reliably across many patients with different anatomies, diseases and clinical needs.

Neuralink’s PRIME study is still an early feasibility study. ClinicalTrials.gov describes it as a first-in-human study evaluating the safety and functionality of the N1 implant and R1 robot in people with tetraparesis or tetraplegia. The study includes long-term monitoring of device- and procedure-related adverse events.

That means there are still many unanswered questions.

How long will the electrodes remain stable?

How will scar tissue affect the signals?

How will the brain respond over years rather than months?

Can the procedure be safely repeated?

What happens if the device needs to be upgraded?

What happens if it needs to be removed?

And perhaps most importantly:

Can the entire process eventually become predictable enough for routine clinical use?

These questions may ultimately be just as important as the number of electrodes or the sophistication of the decoding algorithm.

Neuralink’s bigger challenge: making the brain interface scalable

Neuralink has already demonstrated why BCIs can be medically meaningful.

The company’s first human participant was able to control a computer using neural signals, and the company’s clinical programme has expanded internationally. The current PRIME study is investigating the technology in people with severe motor impairment, while additional trials are exploring other applications, including communication restoration.

But Neuralink’s ambition goes beyond demonstrating that a person can move a cursor with their thoughts.

The real challenge is scalability.

And scalability in medicine has a different meaning from scalability in software.

A software company can deploy an application to a million computers almost instantly.

A medical device that has to be surgically implanted into the human body cannot be scaled in the same way.

Every patient has anatomy.

Every patient has a medical history.

Every brain is slightly different.

Every operation carries risk.

That is why robotics, imaging, automation and surgical simplification could become just as important to the future of BCIs as advances in neuroscience.

Could BCIs eventually become routine medical implants?

It is tempting to jump from Neuralink’s latest milestone to visions of healthy people connecting their brains directly to artificial intelligence.

But medicine is likely to take a more pragmatic route.

The first major applications are much more likely to focus on people with serious neurological disabilities.

A person who has lost the ability to communicate because of paralysis has an entirely different risk-benefit equation from someone who simply wants a faster way to control a computer.

This distinction will matter enormously.

For someone with severe paralysis, an invasive BCI could potentially provide something profound: communication, independence or control over assistive technology.

For a healthy person, the same surgery would need to meet a much higher safety threshold.

In other words, the future of BCIs may arrive first not as a consumer technology, but as a medical technology capable of restoring functions that disease or injury has taken away.

The real breakthrough may be the surgery, not the chip

This is perhaps the most important lesson from Neuralink’s transdural procedure.

We often think about technological breakthroughs as better chips, more electrodes or smarter artificial intelligence.

But sometimes the breakthrough is much more practical.

It is finding a better way to put the technology into the human body.

The dura mater has protected the brain for millions of years. Neuralink is now trying to work through that protective barrier rather than simply removing it.

If the approach proves safe, stable and reproducible, it could help move invasive BCIs one step closer to becoming a scalable medical platform.

That does not mean Neuralink has solved the BCI problem.

It has not.

The technology remains investigational, and long-term clinical evidence will be essential.

But the direction is significant.

The future of brain-computer interfaces may depend on three technologies advancing together:

better electrodes, better artificial intelligence and better surgery.

And perhaps the most important of the three is the one we tend to notice the least.

Because before a computer can understand the brain, medicine first has to find a safe way to reach it.

The bigger question

Neuralink’s transdural procedure represents more than another step in the race to connect brains and computers.

It points towards a more fundamental question:

Can we transform the brain-computer interface from an extraordinary experiment into an ordinary medical procedure?

If the answer eventually becomes yes, the history of BCIs may look very different.

The revolution will not begin when a machine finally understands the human brain.

It may begin when surgeons can safely, precisely and repeatedly build a bridge between the two.