The Doctor Building Real-Life Cyborgs
TL;DR
Phantom Neuro is chasing Neuralink-like control without brain surgery — Connor Glass says Phantom’s implant sits under the skin in a limb or amputation stump, reads muscle signals that are still active after limb loss, and aims for the same “control with your thoughts” outcome with a far simpler, lower-risk procedure.
Muscle is the loud, practical signal; the brain is the whisper in a crowded room — Glass argues muscle signals are roughly 1,000x larger than cortical signals, making them easier to detect and decode than brain activity, with less noise, fewer electrodes, and no neurosurgery required.
Today’s prosthetics mostly fail because the interface is awful, not because the robots are bad — He compares current control schemes to using a laptop with only two buttons: robotic hands from companies like Psionic look impressive, but patients often abandon them because control is clunky, unreliable, heavy, and insurance is a nightmare.
The hype demos are real, but they’re still the best possible moment — Glass is unusually blunt that demos from Phantom, Neuralink, Onward, Meta, or Figure should be viewed as highly curated snapshots, and says true near-human performance in limb control or spinal recovery is more likely a 10-20 year problem than a 2-3 year one.
DARPA built the modern neurotech stack, but commercialization is colliding with medicine and regulation — He says everything from Neuralink’s robot and threads to Phantom’s implants and spinal systems traces back to DARPA funding, while startups now face a strange split: pitch investors on a trillion-dollar future, then pitch insurers on a small, affordable niche device.
Glass thinks transhumanism is coming, starting with medicine and convenience, not sci-fi mind-melds — He’s “pro-cyborg” and sees smart under-skin implants, proactive amputations for diseases like diabetes or rheumatoid damage, and machine control in factories or defense as plausible next steps, while remaining skeptical that broad brain-AI integration is anywhere close.
Summary
The pitch: Neuralink-style control, minus the brain surgery
Ashley Vance opens big — the future merger of humans and machines — and Connor Glass immediately grounds it. Phantom Neuro, he says, is doing “the same thing that Neuralink does” in spirit, but by implanting under the skin in a limb rather than threading electrodes into the brain. The core tradeoff is simple: if you can avoid brain surgery and still get useful control, the risk-reward equation changes dramatically.
How an amputee can still move a robotic hand
Glass walks through the biology in plain English: motor intent starts in the brain, runs down the spinal cord and nerves, and ends in muscle — and every step generates electricity. Phantom reads those signals at the muscle, which still fires even after someone loses a hand or forearm, because the upstream pathways remain intact. That’s why someone like Alex, using Phantom’s surface version, can think about closing a fist and control a robotic hand “as if it’s their natural limb.”
Why muscle beats brain signals for this job
Here Glass gets into the technical heart of the company. Brain signals, he says, are like someone whispering in a crowded room; muscle signals are like someone yelling, with amplitudes close to 1,000 times larger and in a less crowded environment. That makes them easier to decode, easier to scale, and easier to capture with a simple 15-minute local-anesthetic procedure instead of a full neurosurgical event.
The prosthetics problem isn’t the hand — it’s the “two-button laptop” interface
When Vance brings up Alex using a Psionic hand, Glass gives the line that sticks: modern prosthetics are often like having a powerful laptop but only two buttons to operate it. The hardware has gotten better — five-fingered hands, mix-and-match wrists and elbows from different vendors, improvements fueled by the humanoid boom — but patients still abandon them because the control is miserable, the systems are heavy, batteries die, and insurance fights reimbursement. Sexy demos don’t matter if the human-machine interface still sucks.
Demo skepticism, timelines, and the hard truth about recovery
This is the most refreshing stretch of the conversation. Glass says every demo — Phantom’s, Neuralink’s, Onward’s, Meta’s — should be assumed to show the absolute best moment a company could capture, not fake necessarily, but curated. He thinks dramatic outcomes in spinal cord recovery and high-performance limb control are real and possible, but says true human-level function is more of a 10-, 15-, or 20-year arc than something arriving in the next couple years.
From Oklahoma would-be soldier to Hopkins doctor building “cyborg humans”
Glass’s personal path explains the obsession. He grew up in Tulsa, wanted to join the military and do special forces, but recurring shin splints and stress fractures pushed him away from that path and toward surgery. At Johns Hopkins, expecting the “cyborg human” future to already be basically solved from all the DARPA-era demos he’d seen, he instead discovered how far the field still was from anything commercially usable — and decided to build the missing product himself.
DARPA made the field, venture capital distorted the story
Glass is emphatic that DARPA created the modern neurotech landscape: Neuralink, spinal interfaces, Phantom-like systems — all of it. But he also explains the startup double bind brilliantly: to VCs, your tech must be for everyone and worth trillions; to insurers, it must be for almost nobody and cheap enough not to blow up reimbursement budgets. That tension shapes the whole industry, especially when the first users are patients with amputations, paralysis, ALS, or locked-in syndrome rather than mainstream consumers.
The bigger vision: universal control, warfighters, Australia, and being “pro-cyborg”
By the end, Glass zooms back out. He imagines Phantom’s longer-term role as a kind of universal controller for a world full of robots, drones, smart homes, and factory systems — useful first in places with extreme cognitive load, like manufacturing floors or military operations. He’s open about the defense implications, blunt that the U.S. is slower than China and too bogged down by FDA-era red tape, and says that’s one reason Phantom’s first implant trial is happening in Australia. His bottom line is surprisingly measured for someone building cyborg tech: yes, he’s pro-transhumanism and “pro-cyborg,” but no, he does not think direct brain-AI fusion for everybody is around the corner.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
The Retirement Email Isn't a Warning
Model retirements now arrive every few weeks; the config-eval-rehearsal loop turns each deprecation email from a fire drill into an afternoon swap.
Playbook
The Cheapest Model That Passes
OpenRouter lists 400 models behind one API. The fix for choosing isn't a better leaderboard, it's a four-step protocol that ends in a real eval.
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.