Brain-Computer Interfaces & Neurotech: Exploring the Last Frontier

TL;DR

If you only have a moment to spare, here are a few things you should know about BCIs and neurotech:

The human brain – our last great frontier. We may be closer to sending humans to Mars or achieving AGI than fully understanding what happens inside our own minds. With a network of around 86 billion neurons, the brain remains the most complex natural system known to us—perhaps even more complex than any system we've ever created.‍

Brain-computer interfaces (BCIs) are beginning to make a mark. In the past two decades, we have come a step closer to decoding the brain and making the information useful through BCIs, thanks to pioneers like Blackrock Neurotech and recent entrants like Neuralink and Synchron paving the way.‍

Invasive and non-invasive BCIs. Invasive BCIs such as Neuralink's N1 involve surgically implanted electrodes that directly interface with the brain’s cortex. In contrast, non-invasive BCIs capture neural signals through the skin and skull, avoiding surgery. However, the data fidelity of non-invasive BCIs is significantly lower due to signal attenuation.‍

From speech decoding to solving the mental health crisis. The neural data from BCIs is decoded into actions and intents – such as controlling a cursor or generating speech, enabling communication and mobility in paralyzed individuals. BCIs can also work the other way around: by stimulating the brain, BCIs can “input” information into the brain, helping to restore senses, slow down neurodegenerative diseases, or address treatment-resistant depression.‍

However, sensing modalities are not developing fast enough. Non-invasive BCI development is being held back by limited sensing capabilities – most devices still rely on century-old EEG technology. While AI has improved signal decoding, fundamental hardware breakthroughs are needed to achieve high-resolution, practical BCIs.‍

Non-invasive BCIs are already making a mark. Devices, such as Flow’s non-invasive headset, are already targeting consumer-grade applications ranging from enhancing focus and productivity to tackling mental health problems. With the further development of experimental magnetic modalities like OPM and similar technologies, we might see the gap narrowing between non-invasive and invasive BCIs.‍

Invasive systems will remain in the realm of serious medical conditions. While invasive BCIs come with higher risks, they remain the most effective solution for restoring lost function in cases of severe paralysis, epilepsy, and neurodegenerative diseases where the benefits far outweigh the surgical risks.

While we are perhaps decades away from certain applications, the promise of BCIs and neurotech is realizing itself rapidly – and poised to redefine how we interface with technology and the world around us.

...

‍Topic Primer

What are Brain-Computer Interfaces (BCIs)?

Brain-computer interfaces (BCIs), also known as Brain-machine interfaces (BMI), are systems that enable direct communication between the brain and external devices, bypassing traditional pathways like muscles or speech. They work by interpreting neural signals and translating them into actionable commands for computers, prosthetics, or other machines. These technologies are no longer a distant dream—they are already helping people regain mobility, communicate, and interact with the world in ways that were unimaginable just decades ago.

BCIs come in three broad categories:

• Output BCIs: Transferring information from brain activity to a device (e.g., controlling a prosthetic device).
• Input BCIs: Sending information from an external source, like a computer or sensor, directly to the brain.‍
• Closed loop BCIs: A combination of input and output capabilities. For example, decoding brain activity and then using the data to inform what “information” to send back to the brain.‍

How do BCIs work?

At a high level, BCIs capture brain activity through sensors—either invasively (implants) or non-invasively (wearables). These signals are processed by algorithms that decode neural activity and translate it into commands that can control devices or software applications.

Why should we be excited?

BCIs represent a paradigm shift in how we interact with technology and the world around us offering solutions to some of humanity’s most pressing challenges. From restoring mobility and communication for people with severe disabilities to enabling entirely new forms of human-computer interaction, BCIs are creating opportunities for profound societal impact. The market potential spans healthcare, gaming, AR/VR, and beyond—making this one of the most interesting frontiers in technology.

Understanding BCIs

Input BCIs

Input brain-computer interfaces deliver signals directly to the brain to stimulate neural activity, enabling sensory restoration, cognitive enhancement, or new modes of interaction. Unlike output BCIs, which decode brain signals, input BCIs focus on modulating the brain to influence perception or behavior.

Invasive vs. non-invasive input BCIs

Invasive systems, such as those developed by Neuralink and Blackrock Neurotech, use implants to stimulate specific brain regions, targeting applications like restoring vision, hearing, or motor function.

Non-invasive input BCIs use technologies like transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS) to modulate brain activity without surgery. Companies like Flow Neuroscience are developing tDCS devices in a headband form factor for treating depression and enhancing cognitive or motor performance while Neuroelectrics applies these technologies to epilepsy and mental health.

Key applications for input BCIs

1. Sensory Restoration

• Vision - Devices like the failed Second Sight’s Argus II promise to restore partial vision by stimulating the retina or visual cortex. Experimental visual cortex implants bypass damaged eyes entirely, offering hope for individuals with profound vision loss. Artificial retinas and direct neural stimulation hold promise for even greater advancements.
• Hearing - Cochlear implants, the most successful form of input BCI, have restored hearing for millions with profound hearing loss by directly stimulating the auditory nerve.
• Touch - Neural stimulation devices enable amputees or people with spinal cord injuries to feel sensations like pressure or texture through prosthetic devices, improving control and quality of life.

2. Neurostimulation and therapy

• Treating movement disorders - Implanted electrodes deliver targeted electrical pulses to manage conditions like Parkinson’s disease, reducing tremors and improving motor control. These systems are also used to address chronic pain and epilepsy, providing significant symptom relief.
• Managing mental health conditions - Non-invasive stimulation techniques are effective in treating depression and anxiety. By modulating brain activity, these approaches offer alternative options for patients who may not respond to medication or therapy.

3. Augmenting cognitive function

• Enhancement devices - Devices aimed at improving focus, learning, and athletic performance through non-invasive stimulation. For example, Neurosity’s Crown device promises to boost productivity by increasing focus. Experimental invasive BCIs explore memory enhancement or accelerated learning by stimulating key regions like the hippocampus.

Output BCIs

Output brain-computer interfaces are systems that decode neural activity from the brain and translate it into useful information or commands to control external devices. These interfaces primarily focus on motor output (e.g., controlling prosthetics or cursors) or speech decoding, enabling people with disabilities to interact with the world.

Invasive vs. non-invasive output BCIs

Invasive output BCIs involve implanting electrodes that interface directly with neurons, providing high-resolution neural data that enables users to control cursors, robotic arms, and other devices through thought alone(!). For example, Neuralink uses flexible, high-density electrode arrays for precise neural decoding, with initial human trials focused on restoring mobility and communication for paralyzed individuals (see the PRIME study for more details). With its Utah Array system introduced in the early 2000s, Blackrock Neurotech is a pioneer in the space.

However, these systems face significant challenges. Surgical implantation carries inherent risks, including infection and long-term complications. Hardware durability and signal stability over time remain major concerns, as neural tissue can shift or scar, degrading performance. Additionally, invasive BCIs are costly to develop, require complex regulatory approval, and are currently limited to highly specific medical use cases, constraining broader adoption.

Non-invasive output BCIs typically come in a headband or helmet form factor where sensors on the scalp typically sense the electromagnetic fields our brains generate to decode brain signals. Technologies including EEG (electroencephalogram), MEG (magnetoencephalography), fNIRS (functional near-infrared spectroscopy), or fMRI (functional magnetic resonance imaging) are typically used. First demonstrated in 1924 by Hans Berger, EEG is one of the most established BCI technologies. Companies like Emotiv and OpenBCI offer near-consumer-grade EEG devices, primarily used for research and basic control applications.

These systems are safer and more accessible since they require no surgery. However, they face limitations, including low signal resolution, interference caused by the skull and scalp, and slower response times, making them less suitable for applications requiring high precision or speed.

Key applications

1. Brain-controlled prosthetics - Output BCis can enable paralyzed individuals to control robotics limbs with precision. For example, DARPA-funded projects have demonstrated multi-fingered robotic arm control.
2. Interfacing with machines - BCIs translate neural activity into commands for controlling devices like cursors, virtual keyboards – or even wheelchairs and smart home systems.
3. Speech decoding - Systems like the one developed by UCSF decode neural activity for generating text or speech in real-time for people with ALS. Recent breakthroughs include achieving communication rates of up to 62 words per minute—a pace that approaches conversational speed.

Closed loop BCIs

Closed-loop systems combine input and output technologies: they not only record and decode neural activity but also provide real-time feedback to the brain or body. Unlike output BCIs, which simply read neural signals, closed-loop BCIs create a bidirectional interaction, where the system continuously adjusts its output based on brain activity or external responses. This dynamic feedback loop is crucial for applications like adaptive neurostimulation, where a device can modulate brain activity to treat conditions such as epilepsy, Parkinson’s disease, and depression. In motor rehabilitation, closed-loop BCIs can enhance neuroplasticity by reinforcing correct movement patterns, helping stroke patients regain motor control.

100 years of brain-computer interface history

Problem space: A need for better recording modalities and decoding algorithms

Sensing modalities are a fundamental bottleneck    

Decoding brain activity at high resolution is still very very hard. To do it at the single-neuron level, microelectrode arrays or electrocorticography decoders are the current cutting edge. Both of these are invasive. Non-invasive methods provide a safer alternative but lack the precision needed for tasks like speech decoding. Yet, we are not sure whether a single-neuron resolution is needed at all, we might achieve the same results by improving our ability to analyze higher-level, lower-resolution signals, using AI.

Better algorithms alone won’t overcome the limits of current sensing technologies. To bridge the gap between resolution and invasiveness, sensor advancements are key. High-resolution non-invasive methods like OPM-based MEG show promise but remain bulky and expensive. Experimental approaches like ultrasound and nanoparticles are still early-stage and largely speculative. The ultimate goal is near-invasive resolution without major surgery, but that remains a long-term challenge.

Maintaining long-term stability, biocompatibility, and signal quality  

Long-term signal stability is a major challenge for implantable BCIs, as seen in Ian Burkhart’s experience, where infection risks led to device removal despite years of success. Beyond stable neural signals, implants must withstand hardware wear, health fluctuations, environmental changes, and biocompatibility issues. Innovations in materials, like INBRAIN Neuroelectronics' graphene-based implants, offer hope for more durable and flexible solutions.

Capturing the signal through the noise: decoding complexity & data scarcity

Current BCI technology is like trying to communicate using Morse code – slow, inefficient, and requiring extensive effort to encode and decode meaning. To truly unlock BCIs, we need a universal neural language translator – one that can seamlessly interpret the brain’s rich, complex signals in real-time and at high fidelity. However, the complexity of neural patterns and our limited understanding of brain encoding make this a major challenge. Environmental interference, such as Earth's magnetic fields affecting OPM-MEG, further complicates signal clarity.

While AI and foundational models hold promise for improving decoding, they rely on high-quality neural data, which remains scarce due to the limited number of human implants—though trials from Synchron and Neuralink are gradually expanding this dataset. Non-invasive methods and animal models may help, but their ability to accurately translate findings to humans remains uncertain.

Creating an “out of the box” experience: calibration, ease of use & device integration

Most BCI systems currently require extensive manual calibration (referred to as programming) and depend on experts for everyday operation. This is resource-intensive and impractical for widespread adoption. Although AI is being increasingly integrated to make programming and remote configuration more efficient, achieving “out of the box” operation remains a technical and operational hurdle.

Eventually, the ease of setting up and calibrating a BCI needs to resemble that of an everyday consumer device – if it is to achieve mass adoption. Furthermore, the devices must seamlessly interact with various digital platforms and systems. As consumer tech is rapidly evolving, BCI makers must keep up to avoid obsolescence and ensure reliability and longevity.

Bringing down cost: manufacturing complexity & quality assurance

Scaling products like Kernel’s fNIRS-based technology or INBRAIN Neuroelectronic’s graphene-based interface from clinical prototypes to commercially viable products remains a significant hurdle due to complex manufacturing processes where high-yield production is essential for profitability. Due to stringent quality requirements, for example, Prophetic tests its transducers by 3D scanning them in deionized water to measure acoustic pressures and ensure safety, which can take  ~2 hours/transducer. Ensuring scalable, cost-effective manufacturing while maintaining rigorous QA is a key hurdle for the industry.

Cutting through regulatory and reimbursement barriers: data & evidence generation

As Class III medical devices, BCIs must undergo rigorous clinical trials and data collection to gain regulatory and reimbursement approval. Securing insurance coverage requires not just proof of patient benefit but also early engagement with insurers. Companies must work with regulators to define clear outcome measures and build evidence early. Even after demonstrating safety, proving efficacy in a way regulators can easily assess remains a major challenge. Without standardized success criteria, navigating the regulatory landscape remains a key industry hurdle.. 

Building trust with consumers: addressing the sci-fi perception of BCIs

BCIs often carry a sci-fi reputation, leading to misconceptions about their purpose and potential. Industry professionals highlight the importance of framing BCIs as medical devices or prosthetics designed to address real, immediate needs—such as restoring communication or mobility—rather than abstract or futuristic AI experiments where data is misused. Overcoming public fears about implants and demonstrating their safety, practicality, and tangible benefits is a critical communication challenge. Misunderstandings also arise from conflating BCIs with concepts like "mind control", overshadowing their actual function.

Tailwinds (& timing)

The growing integration of AI & ML

AI, particularly deep learning and machine learning, is transforming BCIs by improving the interpretation of complex neural signals. These advancements enable more accurate thought-to-action translation, enhancing precision in brain-controlled interfaces. While AI should not be overhyped, machine learning is driving more robust signal processing, adaptive calibration, and universal decoders that reduce setup time for neurotech devices.

AI has also made breakthroughs in restoring communication for individuals with ALS or stroke-induced speech loss. Models can now decode speech-related neural activity and convert it into real-time synthesized speech or text, with some systems achieving near-natural conversation speeds already in 2023. More recently, in early 2025, Meta AI’s Brain2Qwerty demonstrated a MEG-based decoding system using deep learning that achieved character error rates (CER) as low as 19% for top participants, narrowing the gap between non-invasive and invasive BCIs.

Miniaturization of hardware & material innovation

Hardware components are becoming smaller and more efficient. Smaller implantable devices mean that surgical procedures can be less invasive. This reduces the risk of complications, minimizes scarring, and shortens recovery times for patients requiring neural implants. With new material innovation, invasive BCIs can be designed to integrate seamlessly with neural tissue, reducing the body's immune response and the likelihood of rejection, as well as increasing the longevity of implants.

Advances in microfabrication allow for more electrodes to be placed in a smaller area. This increases the resolution and fidelity of neural recordings and stimulations, enhancing the performance of BCIs. Miniaturized processors can be integrated directly into implants or wearables, allowing for advanced signal processing within the device. This reduces latency and enhances real-time performance, which is critical for applications like prosthetic control.

Increased funding by private and public institutions

Growing public interest & Big Tech dabbling in the space

High-profile companies like Neuralink have sparked widespread excitement about neurotech, boosting public curiosity and driving investment into the field. Similarly, Big Tech is exploring how BCIs can enhance consumer experiences in areas like AR, VR, and gaming. For example, Snap acquired NextMind in 2022 and Meta acquired CTRL-Labs in 2019, and Nvidia’s recently announced collaboration with Synchron promises to combine AI and advanced computing to improve real-time neural signal processing for Synchron’s stentrode device. These signals indicate that neurotech is no longer just a niche academic field—it’s becoming a key area of focus for tech giants shaping the future of human-computer interaction.

Increasing prevalence of mental health problems

The growing prevalence of mental health issues like depression, anxiety, and cognitive decline is creating significant momentum for the neurotech industry. Depression, anxiety, and other mental health disorders have surged globally, exacerbated by stressors such as the COVID-19 pandemic and social isolation. These conditions represent enormous unmet clinical needs that neurotech companies are positioned to address. For example, Kernel’s non-invasive technology that measures hemodynamic changes in the brain can predict treatment responses in depression. This represents a paradigm shift from the trial-and-error approach traditionally used in psychiatry

Companies in the space

We believe…

1. Non-invasive BCIs will target broad, scalable applications

The future of neurotech lies in non-invasive solutions that eliminate surgical risks, making them accessible for both clinical and consumer use. Wearable EEG devices are already emerging for mental health, cognitive assessments, and diagnostics. By integrating into everyday life without specialized expertise, these technologies could scale broadly—but achieving their full potential depends on advances in sensing and decoding.

2. Invasive systems will be constrained to patients with severe medical conditions (for now)

For the foreseeable future, invasive BCIs will remain focused on patients with severe neurological disorders, where their life-changing benefits outweigh the risks and costs of surgery. These systems offer unmatched signal fidelity, enabling applications like continuous speech decoding, complex motor control, and cognitive restoration. Much like traditional medical devices, their high development costs are justified by their transformative impact on critical conditions.

3. Neural decoding of BCIs is likely to experience a “ChatGPT Moment”

A 'ChatGPT moment' for BCIs could come when AI enables real-time, highly accurate neural decoding from non-invasive devices. Just as GPT models suddenly made human-like text generation possible, a breakthrough in neural decoding could allow users to seamlessly control devices through thought—whether typing, navigating interfaces, or operating prosthetics with natural precision. Making this widely accessible through affordable, non-invasive devices would mark a paradigm shift in human-computer interaction.

4. Neurotech will converge with everyday technology

Neurotech will seamlessly integrate with consumer platforms like smartphones, wearables, and smart home devices, enabling users to interact with technology through thought alone. Much like how Oura transformed at-home sleep tracking or smartwatches transformed health tracking, brain-sensing wearables—embedded into familiar form factors like headbands or earpods—will become intuitive tools for productivity, wellness, and immersive experiences.

5. But ethical and privacy concerns will create novel complexities for regulators

The widespread adoption of BCIs will raise unprecedented ethical and privacy concerns. As these devices capture vast amounts of neural data, questions of consent, security, and potential misuse will spark intense societal debates over cognitive liberty—the right to control access to one’s thoughts. Regulators must strike a delicate balance between enabling innovation and protecting individual autonomy, ensuring neural data remains secure while fostering responsible neurotech development.

... and the more Sci-fi predictions

1. BCIs will be able to simulate virtual realities and dreamscapes

Future BCIs could unlock hyper-realistic virtual realities, where users experience lifelike dreamscapes and immersive simulations. Imagine stepping into a book, exploring simulated lifetimes, or even solving problems in dream-induced thought experiments. What once seemed like science fiction may not be far off.

Companies like Prophetic are already working on inducing lucid dreaming using Transcranial Focused Ultrasound (tFUS) to stimulate the Dorsolateral Prefrontal Cortex (dlPFC). Their system reads brain activity via EEG, processes it with an AI-driven neural model, and delivers targeted stimulation to enhance lucid dreaming states.

A future where BCIs generate fully immersive dreamscapes raises deep existential questions: If reality can be simulated at will, how do we define what’s real? At the same time, such advancements could elevate human creativity and problem-solving beyond current limits.

2. Unlocking neural telepathy & the bandwidth for human communication

Advanced BCIs could enable direct brain-to-brain communication, expanding human interaction beyond spoken language. People might share thoughts, emotions, or experiences in real-time, creating a form of 'neural internet' where minds are interconnected.

Early experiments have already demonstrated basic brain-to-brain communication, such as teams collectively playing a game via neural signals. AI-driven neural decoding continues to improve our ability to reconstruct thoughts and intent from brain activity. If these advances continue, BCIs could one day enable richer, more nuanced communication than words allow.

3. And the most sci-fi: enabling the uploading & downloading of consciousness

The ultimate frontier for neurotech is mapping, digitizing, and potentially uploading human consciousness. This could one day enable 'digital immortality,' where minds exist beyond biological limits, transferring into virtual environments, robotic bodies, or even other hosts. Imagine instantly acquiring a new language or reliving historical events by 'downloading' experiences directly into the brain.

While we are far from this reality, foundational research is underway. The brain's 86 billion neurons and 100 trillion synapses form a dynamic network, and capturing both its structure and function is a massive challenge. The Human Connectome Project and similar initiatives have made strides in mapping neural connections at a macroscopic level, but decoding consciousness itself remains a mystery.

What we are excited about

Neurotech and BCIs present a transformative opportunity to redefine how we interact with technology and the world. Building a generational company in this space will likely require some level of vertical integration—much like SpaceX in the space industry—where cutting-edge hardware and advanced neural decoding algorithms work seamlessly together. Neuralink exemplifies the power of rapid iteration, full-stack control, and optimizing cost structures across the supply chain.

Yet, with current sensing and decoding limitations, long-term value will also accrue to those solving these foundational challenges—whether through breakthrough hardware, next-gen software, or entirely new approaches. Who will become the Nvidia or OpenAI of BCIs? Or will Big Tech dominate the space?

If you’re building core neurotech infrastructure, developing game-changing applications, competing with Neuralink, or simply exploring the field, we’d love to connect.

…

Thank you Ian Burkhart, Peter Zhegin, Tue Lehn-Schiøler, Mariska van Steensel, Kurt Haggstrom, Ryan Field, Eric Wollberg, Bálint Várkuti, Mahdi Davari, Mark Yousef for taking the time to share your perspectives and discuss the BCI space with me.

‍

Additional resources:‍

Report: Morgan Stanley’s 2024 report on BCIs
Book: The Battle for Your Brain: Defending the Right to Think Freely in the Age of Neurotechnology by Nita Farahany
Book: Gray Matters: A Biography of Brain Surgery by Theodore Schwartz
Market Analysis: The State of NeuroTech: Unlocking Minds & New Markets by Yuliya Sychikova (DataRoot Labs)