Why Brain-Computer Interfaces Belong in the Future of Work Conversation
“The boundary between human cognition and machine cognition is going to get less clean.”
Roughly once a quarter, I try to interview someone who is living a little further into the future than the rest of us.
This week’s conversation is with Rob Toews, a technology investor and writer focused on AI and brain-computer interfaces, or BCIs.
BCIs are technologies that create a direct connection between the brain and a computer. Some are invasive, like Neuralink, requiring surgery to implant a device into or onto the brain. Others are non-invasive, more like a headset, cap, or wearable that sits outside the skull.
I know. This sounds far-fetched, because it is. The most sophisticated technologies are still largely clinical, and broad commercial adoption is likely more than a decade away.
But a lot of the future-of-work conversation is already operating on a ten- or twenty-year horizon. We are debating which jobs will exist, which skills will matter, what schools should prepare people for, and how much of human knowledge work AI may eventually absorb.
If that is the time horizon, then BCIs probably belong in the conversation. For some reason, they mostly are not.
That omission started to seem strange to me after reading Rob’s work. Many of the people building and backing these technologies do not see BCI as a offshoot of the AI story. They see it as one possible interface layer between human intelligence and machine intelligence. In other words, they are not just trying to build better computers. They are trying to change how humans communicate with computers in the first place.
And interfaces have always shaped work.
The printing press, the telephone, the computer, the internet, the smartphone. Each changed how quickly and richly people could share information, coordinate, and create together.
Our current interface with AI is still primitive. We type into a box. We talk into a microphone. We wait for a response. It may feel magical compared to where we were a few years ago, but it is still a lossy way to transmit thought.
BCI points toward a more radical possibility. What if the bottleneck is no longer language? What if the interface between humans and computers moves closer to thought itself? What if knowledge workers, fifteen years from now, can share context with coworkers and AI systems in ways that feel closer to telepathy than typing?
Most of the AI and work conversation is about substitution. What will AI do instead of humans? Which tasks will disappear? Which jobs are exposed?
BCI opens up a different line of inquiry. What could humans and AI do together if the interface between them became dramatically more powerful?
That question runs straight into education.
If knowledge can flow more seamlessly between a person, an AI system, and eventually another person, then we have to rethink what learning is for. What needs to be encoded in the brain? What can be accessed from outside it? What capacities become more valuable when information itself becomes easier to summon?
My bet is that discernment, agency, judgment, motivation, and the ability to know what to trust become more important, not less. The optimistic version of this future is expanded human capability: faster communication, new forms of creativity, better support for focus, sleep, mental health, and cognition. The darker version is surveillance, influence, and control moving closer to attention, judgment, and thought itself.
If you’re anything like me, this conversation will leave you with the sense that the future is probably stranger than any of us can imagine.
Allison
On The Inevitable Merger of Human and Machine Intelligences
ALLISON: What’s something that you’ve changed your mind about in the past year?
ROB: At the highest level, the world of brain-computer interfaces (BCI) divides into two camps: invasive and non-invasive. Invasive methods require surgery to implant a chip into or onto the brain — that’s what Neuralink, for instance, is building. Non-invasive methods sit outside the skull, like a wearable headband or cap. That divide has been fundamental to the direction of BCI development.
In thinking about which approach will win out, I’ve come to see these not as competitive approaches but as two players in a broader ecosystem. BCI is not going to be one product or technology — it’s a sprawling domain of applications. Some use cases will favor non-invasive methods; others will call for invasive approaches, despite their added complexity, because they unlock the most sophisticated BCI capabilities. Both will be successful and prevalent, but they will serve different purposes.
ALLISON: You’ve written that it’s not possible to understand the long-term future of AI without understanding brain computer interfaces. That’s an ambitious claim. Can you walk us through it, and how it fits into the bigger conversation around AI and the future of work?
ROB: The simplest way to say it is that BCIs may become the interface layer between human intelligence and machine intelligence.
Right now, our interface with AI is still incredibly clunky. We type into a box, talk into a microphone, and wait for a response. That is a very low-bandwidth way for humans and machines to work together.
BCIs point toward a world where that interface gets much closer to thought itself. At first, that may mean restoring movement or communication for people with paralysis. Over time, it could mean faster communication, richer collaboration with machines, new forms of creativity, and eventually a much tighter coupling between biological and artificial intelligence.
That is relevant for the future of work because most of the current AI conversation is about substitution. Which tasks will AI do instead of humans? BCI opens up a different question. How might humans work with AI if the interface between us and machines became dramatically more powerful?
I do not think that means everyone will have a chip in their brain. Some applications will be invasive because they require the highest possible fidelity. Others will be non-invasive, more like a headset or wearable, because that will be good enough for the use case. But across both categories, the deeper point is the same. The boundary between human cognition and machine cognition is going to get less clean.
That is why I do not think BCI is a niche neuroscience topic. It is one of the places where the future of human intelligence and the future of machine intelligence meet.
ALLISON: Is this trajectory inevitable?
ROB: Yes. You could interpret the history of communications technologies as an ongoing progression of the ability to transmit information faster and more reliably, from the printing press onwards, the positive result of which are advances in science, healthcare, education, and so forth. The apex of that progression is BCI; it will enable higher bandwidth communication and more high-powered cognitive processing than would otherwise be the case. And humans are a curious species, and there are many very talented innovators and technologists who are pushing this tech forward, whether we think it’s a good idea or not.
ALLISON: Do you think it’s a good thing? Are you excited about this future?
ROB: I am excited about it, and I don’t think it’s necessarily utopian nor dystopian. It will be complex and different. But I’m optimistic that humans will continue to flourish and play a central role in this world. We’ve incorporated many technologies into our lives and our internal experiences.
I also think there will be quite a lot of social uplift that we will be able to objectively measure in terms of things like life expectancy, mental health, education levels, and GDP per capita.
On Introducing BCIs to the Future of Work
ALLISON: The dialogue about how AI is changing the future of work isn’t yet considering BCIs. Could you orient us to what work looks like when BCIs are commercially available, prolific and in a variety of form factors?
ROB: You’re right that people aren’t widely talking about this yet — BCIs are about two years from being a much more widely appreciated technology. On the invasive side, regulatory approvals at patient scale are 2029–2031 at the earliest, barring clinical trial delays. The initial focus will be on clinical populations and medical use cases, and broader access to sophisticated invasive BCI is still several years beyond that. AI Agents, by contrast, are capable and accessible right now, which is why they’re front and center in the conversation about work.
For the knowledge worker, BCI presents a dramatically more powerful interface between humans and machines. That could take many forms, but one shorthand I like: it will make telepathy an everyday thing. There is always a slippage between our thoughts and what gets expressed in natural language; what might it feel like to express the meaning fully with other people with the medium of language? That offers a much higher bandwidth than simply using words. That would be incredibly compelling in terms of how quickly we can communicate with one another and with computers, and would fundamentally change how we work.
AI will automate everything that knowledge workers do today. The horizon breaks in two directions: a post-work world, or the discovery of new kinds of work — new things that matter, new games to play with one another. The latter seems more likely to me.
ALLISON: What does BCI at work look like in the near-term for knowledge workers?
ROB: The nearest-term applications for knowledge workers will be BCI tools that improve focus, sleep, and brain/mental health. Let me break down how those work.
An interesting distinction within BCI development is between tools that read from the brain and tools that write to it. The first category senses your neural patterns and decodes what you’re thinking, what you want to say, or what you want to move — then lets you express yourself via thought. Telepathy and telekinesis both involve reading from the brain.
The second category provides inputs that intervene in your thoughts, mood, mental state, and cognition. A well-established body of peer-reviewed literature shows that electrical stimulation in specific parts of the brain — at precise intensities and frequencies — can improve focus, decrease stress, reduce depression, and improve sleep. A number of these products are already on the market.
Thought-to-text will be a near-term innovation for knowledge work — perhaps two to three years out. My hypothesis is that the first company to crack 99%+ accuracy will take the invasive approach, though signals suggest non-invasive thought-to-text will eventually be possible as well.
ALLISON: When will it be the default technology mediating knowledge economy work?
Rob: I think it will be very much mainstream before 2040, because I think folks will be at a disadvantage of not using this higher bandwidth interface.
ALLISON: So based on your prediction, kids in elementary school today may never experience a world of work where they’re not in some way merged with one another and with AI intelligence.
ROB: I think that’s right.
ALLISON: Eerie.
On Today’s BCI Applications for Bolstering Human Cognition and Learning
ALLISON: What does the landscape of BCI research into cognition look like, especially when it comes to things like executive function, focus, and memory? What makes you most excited about how BCI will push human cognitive capability forward?
ROB: It’s worth repeating that we still don’t know enough about the brain to have a clear line of sight into cognitive functions. We know vision is in the occipital lobe and hearing in the parietal lobes, but we lack a detailed map of the subregions and neuron clusters responsible for focus, memory, and so forth. Most cognitive tasks are distributed across the brain, and tools like fMRI can’t always tell us exactly where. As research advances, I do think BCI will unlock more of those capabilities.
That said, we know enough, and with enough precision, about how the brain’s electrical signals shape subjective experience to build products that support focus and attention. Several companies have done exactly that: Neurode in Australia and Flow Neuroscience in the UK both sell consumer headbands online today. Non-invasive products like these require a calibration period, but as they learn how information moves through your brain, they begin sending electrical pulses — tuned precisely for you — to nudge you toward greater focus and attentiveness. We may not know the exact neuron to target, but we know how to make the intervention.
These products work well. I’m a strong advocate for the sleep tools in particular — an incredible use case that I expect to see wide adoption. Efficacy studies from several companies show their tools are four to five times more effective than melatonin and 50% more effective than Xanax. Almost everyone wants more and better sleep. This feels like the next great consumer wearable; these kinds of non-invasive BCI products will be widespread in 18-24 months.
ALLISON: Get theoretical for us for a moment: What role does knowledge acquisition (in education) play in a world where you have a near seamless and instantaneous flow of data between an artificial intelligence in your brain and between one brain and another brain? For me, the BCI puts the whole notion of what needs to be encoded in my brain in question.
ROB: In a very real sense, as we discussed earlier, artificial intelligence and human intelligence will start to meld together and merge. It may be hard to crisply distinguish between the two. So when you’re thinking about a topic, or brainstorming an answer to a question, information and data from the AI will be part of that cognitive process with you — you may not be able to discern what knowledge is yours and what’s stored in the cloud. To which I ask: does it matter? At what point is it relevant? A core supposition of this thought exercise is to assume that AI will become part of our identity and cognition.
But because we are human, we will always value other humans in and of themselves, for no other reason than that we value humanness. A superintelligence can’t solve the entire class of activity that is about connection, physical endeavor, and relationship. That dimension of our experience will never not be relevant.
ALLISON: I want to push on that, because I think a lot of people would say yes, it absolutely matters.
Not because memorization should remain the center of education. I agree that instant access to knowledge changes what we need to encode in our brains. But it matters a lot whether we can tell the difference between our own thoughts, another person’s thoughts, and machine-generated information.
If that boundary gets blurry, discernment becomes more important, not less. The question is how we know what to trust, how we notice when we are being shaped or steered, and how we preserve real sovereignty over our own attention and judgment.
Because the darker version of this is not just AI becoming part of cognition. It is the systems that help us think becoming mechanisms for surveillance, influence, and control. As the conversation around BCI x AI continues, it feels critical that we define really clearly how this technology expands human agency, rather than colonizes it.
One Small Signal
ALLISON: What’s one small signal in the world right now to which we should be paying more attention?
ROB: On the more tech-forward front, I don’t think there is enough widespread appreciation and focus on how transformative BCI is going to be to AI’s progress more generally. And as I’ve suggested, I think that will change in the next year or two.
ALLISON: What’s going to drive this change in perception?
ROB: Most people still think of this technology as science fiction. Elon Musk’s Neuralink has been around for years and it’s commanded attention and interest, but I don’t think people appreciate how near-term this really is. In the next 10 years, invasive BCI users will number in the many thousands, perhaps millions.
Within BCI, ultrasound is emerging as a massively promising application, especially for non-invasive approaches. It outperforms existing modalities like EEG and fNIRS because it penetrates far deeper into the brain than any other non-invasive technology, and with much greater precision — down to the sub-millimeter scale. Crucially, you can both read and write with it, which sets it apart from other non-invasive approaches. It’s still nascent, but with major players entering the space — Sam Altman has a new BCI startup pursuing ultrasound; Nudge is, I think, one of the most interesting — I believe we’re five years from ultrasound becoming the dominant non-invasive BCI modality.
ALLISON: People are used to ultrasound imaging, so I think the reading capability is a little easier to understand: ultrasound waves can abstract data and make us identify and make sense of activity patterns. The writing capability is perhaps less obvious: my understanding is that you can actually use ultrasound to neuromodulate the brain state? Is that the thrust of it?
ROB: Exactly. Focused ultrasound works by sending out multiple ultrasound waves timed so their peaks converge at exactly the same point. When they all align at a single location in the brain, the combined energy is intense enough to mechanically alter how neurons function — modulating brain activity at that precise spot. This ability to concentrate energy with that kind of accuracy was a major breakthrough that made ultrasound a genuinely powerful neuromodulation technique.
ALLISON: Thanks so much, Rob, for giving our readers this intro into BCI and how it works; I think it’s been helpful for me to learn a little more about the technology behind the idea, much of which, like ultrasound and wearable devices, are things we already have familiarly and fluency with. As BCI moves from science fiction to store shelves, the future is about to get weird.