We scan new podcasts and send you the top 5 insights daily.
Matias says old medical devices were single-indication assets, but brain-computer interfaces are becoming horizontal platforms. Miniaturization plus better algorithms enable closed-loop read-and-stimulate systems, and neurotech's circuit-level interface makes many psychiatric applications plausible. That shift is why he is excited about devices despite their historical reputation.
The performance ceiling for non-invasive Brain-Computer Interfaces (BCIs) is rising dramatically, not from better sensors, but from advanced AI. New models can extract high-fidelity signals from noisy data collected outside the skull, potentially making surgical implants like Neuralink unnecessary for sophisticated use cases.
Matias says his time at NVIDIA gave him an insider view of AI progress—more compute, better data, better models—converging with neurobiology. That visibility helped convince him a venture fund focused on brain and mental health could work. He calls it one of the lucky breaks behind SciMed's launch.
The company's next product will provide objective brain state data, much like a CGM provides constant glucose readings. This allows for data-driven mental health treatment, moving beyond subjective checklists and enabling closed-loop therapies with neuromodulators, fundamentally changing diagnostics and care.
Frontier AI labs like Meta, OpenAI, and Anthropic are ultimately tracking toward becoming biotechnology platforms. While the public focuses on large language models, the real competitive frontier involves dynamically reading and writing to human brain circuits non-invasively. By pairing AI with modalities like focused ultrasound, these companies aim to modulate motivation, alertness, and learning in real time by titrating activity in specific subcortical regions such as the nucleus accumbens.
There's a spectrum of BCI applications. One end offers "hand substitutes," like silent speech for app control, which may offer only marginal utility. The more transformative end focuses on restoring core sensory channels like vision and hearing, which fundamentally augments the human experience.
Brain interfaces do not initially comprehend the user's motor intention signals upon implantation. Instead, software algorithms identify firing patterns during guided tasks while the human brain simultaneously learns the system's software model, creating a co-adaptive feedback loop that stabilizes daily cursor control.
While companies like Neuralink popularize assistive BCIs for controlling external devices, a different segment is focused on therapeutic applications. Companies like InBrain aim not to control computers but to use high-resolution interfaces to directly heal or modulate neural circuits for treating diseases.
Treating the brain as a computational device allows for an engineering-driven approach to solving neurological issues. This yields more predictable results and faster progress compared to the "random walk" of small molecule drug discovery, which often ends in failure after years of research.
Unlike conventional deep brain stimulation that delivers a constant current, adaptive DBS uses a bidirectional system. It senses neural activity, compares it to a threshold, and adjusts the stimulation up or down accordingly, creating a personalized therapy that responds to the patient's real-time brain state.
The core technology of detecting "intent" is viewed as a platform. Once the implant is in place for stroke recovery, it can be trained to detect cognitive lapses and provide real-time prompts, creating a system to assist with conditions like dementia or MCI.