There is a moment in Neuralink’s video about Terry that makes the whole story feel very small.
Not small in importance. Small enough to understand.
Terry has ALS. Speaking has become increasingly difficult. Near the beginning, he tries to tell an automated phone system that he needs help with a claim. The system doesn’t understand him.
It asks him to try again.
We have all been frustrated by a machine that won’t listen. But imagine knowing what you want to say and finding that the way you have said things your entire life no longer works reliably.
Imagine that happening with your clients. Your friends. Your wife.
Terry’s wife describes a man who loves to talk. Someone who had recently made partner and was enjoying his career. Then communicating at work became harder. Eventually, he went on long-term disability and retired.
Before the demonstration, Terry explains what he hopes the technology will make possible.
“I hope this will allow me to be a part of the conversation again.”
That is the ambition I want us to pay attention to.
Not becoming superhuman.
Being able to join in.
The person has something to say
When we talk about AI, we often ask what a machine can produce.
Can it write an essay? Build an application? Solve a difficult problem?
This story turns that question around.
What if the valuable thing already exists in a person, and technology helps it reach someone else?
Terry has things he wants to say. The challenge shown in the video is getting those words out in a form other people can understand.
Neuralink is studying a possible new route. Its VOICE trial investigates an implanted brain-computer interface that could help people with severe speech impairment turn speech-related neural activity into text or synthesized speech.
In the video, a researcher explains that Terry’s implant is positioned in a region involved in producing speech. Terry attempts sentences displayed on a screen. The team collects the corresponding signals and uses them to build a decoder.
The idea is approachable even if the engineering is not. The system learns relationships between recorded brain activity and the speech Terry is trying to produce.
Later, the computer speaks.
The voice comes through a speaker. The purpose comes from him.
The achievement is not that a computer has found something to say. It is that a person has another way to say it.
A conversation is more than its words
Think about the last conversation you enjoyed.
You probably didn’t take orderly turns delivering complete statements. Someone interrupted. Someone laughed before a sentence was finished. A small detail sent everyone in a different direction.
Perhaps you made a joke that only worked because you said it at exactly the right moment.
That is part of what makes conversation feel alive.
Now imagine having a response ready while everyone else moves on. By the time you can express it, the opening has passed.
A communication tool can convey the correct words and still make participation difficult if using it demands too much time or effort. That is why I find this research so compelling. The opportunity is larger than producing sentences. It is helping people take part more comfortably, with more control over the timing and shape of what they say.
There are already important ways to support communication. Speech-generating devices, eye-gaze systems, communication boards, and voice banking help people express themselves today. The ALS Association describes these options as part of individualized communication support.
Brain implants should not make us overlook those tools or the people who rely on them. An invasive technology will not be the right choice for everyone.
But another useful option could matter enormously to someone whose existing options do not meet their needs.
The goal should be more freedom to communicate, not a particular piece of hardware.
What the demonstration does and does not establish
It is possible to feel the significance of Terry’s experience without pretending every hard problem has been solved.
The video presents one participant’s story. It shows training, an interface, and examples of communication. It does not, by itself, tell us how reliably the system works across an entire day, how often it makes mistakes, or what someone else should expect.
This is an experimental assistive technology. It is not a cure for ALS, and producing speech through a computer is not the same as restoring the biological ability to speak.
Neuralink explicitly states that its devices remain investigational and have not been approved by the FDA or other regulatory authorities.
The broader field offers reasons to take the possibility seriously. In 2025, UC Davis researchers reported a different experimental system that allowed one participant with ALS to generate audible speech from neural activity, including changes in intonation and simple melodies. Their research report also describes its limitations: listeners understood almost 60 percent of the synthesized words, and the findings needed to be replicated in more people.
That is not the same device or the same result as Terry’s. It is evidence that multiple teams are making progress on a difficult problem, with substantial work still ahead.
I find that more encouraging than a claim that the problem is finished.
We have something concrete to improve.
The breakthrough still has to survive an ordinary day
There is another part of this story that feels familiar to anyone who builds software.
Getting a system to work is not the same as making it easy to live with.
On October 29, Neuralink published an engineering update about a 72-hour visit to a VOICE participant’s home. Its team described prototyping automatic speech-start detection, faster correction of decoding errors, and integration with phone calls.
These are company-reported prototypes, not proof of dependable performance for everyone. But the problems they address are revealing.
How do you begin speaking?
What happens when the decoder gets a word wrong?
Can you use it without navigating through several screens?
Those questions might sound less impressive than recording signals from the brain. To the person trying to get through a conversation, they could determine whether the technology is useful.
Imagine needing to correct a simple sentence while the person on the other end of a call assumes you have finished.
Imagine having to choose between sending an imperfect message quickly and losing the moment while you repair it.
The details are not separate from the breakthrough. They are how the breakthrough becomes part of a life.
In The Software You Wish Existed, I wrote about the knowledge held by people closest to a problem. Here, that knowledge is indispensable.
The participant is not just someone engineers test a product on. He knows things about the experience that they cannot learn by looking at a dashboard.
Whose words are they
There is a responsibility that comes with building a system that speaks for someone.
It needs to preserve that person’s meaning.
A sentence can sound perfectly natural and still be wrong. An incorrect word can change a preference, an instruction, or an answer to an important question.
As these systems develop, I would want to see more than measures of speed. How easily can someone notice a mistake? Correct it? Stop an utterance? Decide when the system is allowed to listen and when it should remain silent?
Helping someone communicate should not mean taking control of the conversation away from them.
The same applies to privacy. Neural data should not become something a person has to surrender without meaningful limits in exchange for being heard. We should ask who can access it, how it is used, and what choices remain with the person producing it.
And we should resist treating a speech-decoding demonstration as evidence that a machine can read everything in someone’s mind. That is not what this video establishes.
These questions do not diminish the opportunity. They define the kind of opportunity worth pursuing.
The point is to give someone more agency, including the freedom to say nothing at all.
A future measured in ordinary moments
In The Age of Imagineering, I wrote about the shrinking distance between an idea and something we can make.
Here, that distance becomes unusually personal.
It is the distance between wanting to say something and another person hearing it.
I think we will become much better at reducing that distance for people who face barriers to communication. That is my expectation, not a promise about when a particular device will be available or who it will help.
There is still surgery to consider. Safety and long-term reliability to establish. Training, maintenance, and support to provide. If these technologies become approved treatments, affordability and access will matter as much as the demonstration.
A device that works only with exceptional support is a different achievement from one someone can rely on at home.
But I do not need to imagine a world in which everyone has a brain implant to see why this work matters.
I can imagine a person ordering their own coffee.
Answering a question before someone answers for them.
Calling a friend because they feel like talking.
Having an argument. Making a terrible joke. Changing their mind halfway through a sentence.
Those possibilities may look ordinary from the outside. That is precisely why they matter.
Three words
Near the end of Terry’s video, the conversation turns to what the system is doing. Words are appearing. People react.
Then comes a short sentence.
“I love you.”
I keep coming back to that.
We talk about the future in terms of intelligence, computing power, and capabilities. All of those things matter. They are part of how we get somewhere new.
But they are not, on their own, the reason to go.
Here is a reason: someone has something they want to say to a person they love, and we may be getting better at helping those words arrive.
Terry was already a whole person before the implant. His relationships, his humor, and his value did not depend on a machine understanding him.
The possibility is that the technology can give him more room to express what was already his.
We should build a future in which fewer people have to fight to be heard.
Sources reviewed October 31, 2026. Terry’s experience is described in Neuralink’s video and the supplied transcript. Research findings and company demonstrations are distinguished above from my expectations about the future.
AI engineer, builder, and author of The Age of Imagineering.