Built On A Bad Bet​

A brain-computer interface (BCI) captures neural activity and translates it into control signals for an external device. The most familiar example is a patient with paralysis using their thoughts to move a robotic arm, with the system reading intention and converting it into action in real time.​

The industry has long accepted who this technology is for: surgical candidates with specific diagnoses and the means to afford invasive procedures. Scalability was never part of the design goal for BCI. The tech got treated as a biological necessity, something baked into the nature of the problem that was made early and never revisited.

Today, only a few hundred people have received any form of implanted neural interface, which is a tiny fraction of those living with conditions that BCI could significantly help.​ Meanwhile, cortical electrode arrays, the gold standard for high-resolution signal capture, cover less than a square centimeter of brain surface. The field has spent decades developing a tool that, by design, can only reach a small portion of the people who need it.​

That’s the actual problem, not the advancements in electrode materials, compute or regulatory pathways. It was the assumption that BCI wasn’t necessary for all, calcified into benchmarks, funding priorities and research agendas, all reinforcing each other and built on a premise nobody was questioning.​

When a field optimizes around a flawed premise for long enough, you don’t get incremental progress. You get a very sophisticated dead end.

​If you’re pursuing an approach that challenges accepted constraints, validate it with experts in adjacent fields, not just your own. This helps distinguish between what is truly impossible and what is simply untested.​

For example, at Subsense, we looked beyond the usual voices in neurotech and BCI, spending countless hours with world-leading experts in nanotechnology, physics, chemistry and beyond. We needed fresh perspectives from unexpected places. They all agreed it was going to be a challenging endeavor, but nobody said it was impossible. And that’s the validation I needed.

The lesson learned: Engage external stakeholders before you feel ready.

These conversations can shape how you invest early. If existing tools were built around outdated assumptions, adopting them may lock you into the same limitations. In our case, we couldn’t rely on off-the-shelf neurotech hardware. We had to put resources into building customized systems tailored to our approach. It was a higher upfront R&D cost, but a clearer path to long-term differentiation.

The discussions also emphasized real-world use and human need, forcing us to map our experiments around them. If our approach couldn’t connect to a meaningful application, it wasn’t prioritized. ​

What Breaking Out Of It Looks Like​​

At Subsense, the founding premise was simple: We will reach a population orders of magnitude larger than current BCI paradigms allow. If your goal is broad access, make that a design constraint from day one. Incremental improvements rarely lead to exponential reach without rethinking the underlying model.​

The team reflected that. We didn’t hire primarily from neuroscience or engineering. From day one, we brought in people from materials science, data engineering, nanotech and product design, because the problem lived at the intersection of those fields, not inside any one of them. Interdisciplinary hiring helped us avoid months of misaligned experiments, where, for example, the nanoparticles may have worked in isolation but not in neural environments. When a materials scientist and a neuroscientist disagree about what’s possible, that friction is where the useful thinking happens.

Through these diverse perspectives and relentless iteration, we found what was achievable, the hurdles we’d face with a nonsurgical approach and the potential constraints of real-world adoption. The trade-off was a harder path up front but the chance to build something far more impactful in the long term.​​

Accelerating With AI

AI has drastically changed the life sciences and deep tech industries, and for BCI, it changes the math on access.​

The limiting factor in scaling BCI has never been compute or ambition. It's been signal resolution: the ability to capture meaningful neural data without requiring surgical implants in every patient. AI shifts that ceiling. Feed the right neural data into well-trained models, and you can extract signal fidelity from noninvasive or minimally invasive hardware that would have been dismissed as noise a decade ago. That's not a modest improvement but rather the difference between a technology that serves tens of thousands and one that could serve millions.​

But the operative phrase is the right data. Most existing BCI datasets were generated from a narrow patient population, using hardware built around the assumption that broad access wasn't the goal. Training on that data encodes the same constraints the field is trying to escape. If you are building a BCI company and using AI to help scale access, you need to design experiments around the signals you expect to build the exact datasets that make that possible, from novel hardware, with a broader population, on purpose.

AI accelerates the right bets. The work is making sure you're placing them.

The Real Bottleneck

The patients who stand to benefit most from BCI are largely unreached. The biology was never the barrier; the field decided, early and without much debate, who this technology was for and then built everything else around that assumption.​

That decision is not permanent. It is not technical. And it is not someone else's problem to reverse.

​The question for everyone working in this space is simple: Are you optimizing the existing framework, or are you dismantling the assumption underneath it?

One of those paths leads to better tools for the same small population. The other leads to something that actually changes the scale of who gets helped. The field is capable of the latter. It just has to choose it.​​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?