The Surgeon Building Real-Life Cyborgs
TL;DR
Phantom Neuro is doing Neuralink-like control without brain surgery — Connor Glass says their implant sits under the skin in a limb or amputation stump, reads muscle signals that are still present after limb loss, and can drive devices like a Psionic robotic hand through a much simpler 15-minute procedure.
Muscle is a cleaner signal than cortex for control — Glass argues muscle activity is roughly 1,000x larger in amplitude than brain signals, making it easier to decode reliably than whisper-quiet cortical activity in a crowded, noisy neural environment.
The real bottleneck in prosthetics isn’t the robot hand — it’s the interface — today’s users often get stuck with the equivalent of “two buttons” to control a very sophisticated machine, which is why many people abandon advanced prosthetics despite the hardware getting better.
Most flashy neurotech demos are the best-case moment, not everyday reality — Glass repeatedly warns that demos from Phantom, Neuralink, Onward, Meta, or humanoid robotics should be viewed skeptically unless they work live and unscripted, because this is still a brutally hard engineering problem.
DARPA created much of the modern BCI field, but commercialization is colliding with medicine and insurance — the foundational research behind Neuralink-style systems, spinal implants, and Phantom’s work was heavily defense-funded, yet startups now have to tell VCs the tech is for everyone while telling insurers it’s for a tiny, affordable patient group.
Glass thinks human-machine merging is inevitable, but mass-market brain implants are far off — he’s openly “pro-cyborg,” sees under-skin implants as a more plausible consumer path than invasive BCIs, and says true broad consumer brain interfaces are likely a 10- to 20-year challenge, not a near-term product.
The Breakdown
Neuralink, but in the limb instead of the brain
Ashley Vance opens by framing Phantom Neuro as part of the bigger “future of humans” story, then Connor Glass gives the core pitch: they’re trying to let people control machines with intention, but without opening the skull. Instead of going into the cortex like Neuralink, Phantom’s implant goes under the skin in a limb or amputation stump — a much easier risk-reward proposition.
How an amputee still has the signals to move a robotic hand
Glass explains the biology cleanly: the command starts in the motor cortex, travels down the spinal cord, through nerves, and into muscle — and even after limb loss, much of that upstream pathway still works. So when an amputee thinks “close my fist,” residual muscles still fire, and Phantom reads that electrical activity off the muscle surface rather than digging into the brain.
Why muscle beats brain for decoding control
His big analogy is that brain signals are like someone whispering in a crowded room, while muscle signals are like someone yelling. Because muscle signals are far larger and there are fewer sources to sort through than in cortex, Phantom thinks it can get more robust control with less complexity, less surgery, and no need for a neurosurgeon or general anesthesia.
The external demos are impressive — but the implant is the real point
Vance brings up Alex, the amputee they filmed in Austin, using Phantom’s surface version to manipulate a robotic hand and wrist assembled from multiple companies, including Psionic. Glass says they’ve worked with “tens” of patients on this non-implanted setup, but the first implanted clinical trial is scheduled for early this year, where the payoff should be a much more stable signal than sweat-prone, movement-sensitive wearables can provide.
Why prosthetic hands disappoint people despite all the cool hardware
Glass is blunt here: the robots aren’t the main problem, the human-machine interface is. He compares current prosthetic control to trying to run a laptop with only two buttons — the machine is capable, but the input is so bad that users often abandon expensive robotic prosthetics and go back to cosmetic limbs or nothing at all.
The demo skepticism section
This is one of the sharpest parts of the conversation. Glass says any demo from Figure, Neuralink, Phantom, Onward, or anyone else is usually the best possible moment, not average performance, and people should stay skeptical unless they see it work live, onstage, in real time — like when Meta’s own neural wristband demo glitched despite all its resources.
Connor Glass’s path: failed military dream to surgeon to founder
Glass grew up in Tulsa, wanted to do special forces, got derailed by recurring shin splints and stress fractures, and redirected into surgery because it offered immediate impact. At Johns Hopkins, he arrived thinking the “cyborg human” future was basically ready, then discovered the gap between polished lab demos and commercially usable systems — which is what pushed him to found Phantom Neuro in 2019.
DARPA, warfighters, Australia, and the bigger transhumanist vision
Glass says modern neurotech largely spun out of DARPA’s billions in funding for injured soldiers and high-stakes human-machine control, and he’s candid that the same systems for disability can later become military or consumer interfaces. He also explains why Phantom is running its first implant trial in Australia instead of the U.S. — faster approvals and a 43% R&D tax credit — then ends on his broader belief that humans will merge with machines, just probably first through safer under-skin implants rather than near-term mass-market brain surgery.