A man who can no longer move or speak sits in his living room. He thinks about what he wants to say, and the screen fills with a short list of sentences an AI thinks he might want. He picks one — not with a finger, not with his voice, but with a single thought. The message goes out.
For people with advanced ALS (肌萎缩侧索硬化) or locked-in syndrome, this is not a science demo. It is the daily shape of a partnership between Synchron, a New York neurotechnology company, and OpenAI that went live in 2024.
Key takeaways
- Synchron's Stentrode is implanted through a blood vessel — no open-brain surgery — and sits against the motor cortex to read movement intent.
- In July 2024 Synchron added a chat feature powered by OpenAI's generative models, letting severely paralyzed users text at conversational speed.
- The system does not share the user's brain data with OpenAI, according to Synchron's own announcement.
- The harder problem is no longer "can we read the signal" but "can we predict what the person wants to say next."
The implant that rides a blood vessel
Most brain-computer interface (脑机接口) coverage assumes a surgeon opens the skull. Synchron took a different road. Its Stentrode is a small mesh electrode array mounted on a stent, delivered through the jugular vein and parked against a blood vessel on the surface of the motor cortex. There is no craniotomy, no robot, and no penetration of brain tissue. The company calls this category an "interventional" BCI, and the pitch is scale: a procedure a vascular team already knows how to do.
Once in place, the Stentrode picks up motor intent and sends it wirelessly to a small receiver, which forwards it to a decoding app. The original use case was simple point-and-click — move a cursor, open an app, control a smart-home device. That alone was a meaningful win for someone with no upper-limb mobility.
When the brain meets a language model
The jump came in July 2024, when Synchron announced a chat feature built on OpenAI's generative AI. The idea is unglamorous but powerful: typing is expensive when you can only click one option at a time. A language model watches the context — who you are talking to, what the conversation is about, even the user's stated emotion — and proposes a handful of replies. The user reviews them with the same thought-driven click used to move the cursor, and selects one.
Synchron demonstrated this with "Mark," a person living with ALS (肌萎缩侧索硬化) using the company's BCI platform in its COMMAND early feasibility study in the United States. In the announcement, Mark said that as his disease progresses he expects to lose the ability to communicate, and that the technology gives him a way to keep connecting with loved ones. The company's CEO and founder, Tom Oxley, framed it bluntly: brain-computer interfaces (脑机接口) preserve "a fundamental human right: the freedom of expression."
One detail matters for trust. Synchron stated in its own press release that it does not share the BCI user's brain data with OpenAI. The model sees prompts and context, not raw neural recordings.
What "hands-free" actually means here
"Hands-free" in a phone ad means Bluetooth. Here it means something stricter: no voice, no eye tracker, no physical switch. The user generates a small number of candidate actions, and the AI narrows the space of possible sentences so that each thought-click carries more weight. Synchron says the result lets users text "at conversational speed" rather than composing character by character.
This is where the language model earns its place. A paralyzed user does not need the AI to "know" them — they need it to guess well enough that a thought-click is faster than any alternative. When the predictions are contextual, the gap between intent and expression shrinks.
The bigger bet: an AI that reads intent
Synchron has not stopped at chat. In 2024 it publicly demonstrated a patient using the Stentrode with an Apple Vision Pro headset to play Solitaire, watch Apple TV, and send messages hands-free. In 2025 Apple introduced a BCI Human Interface Device protocol that treats neural input as a native control method for iPhone, iPad, and Vision Pro, and Synchron said its device would be the first to integrate. Later that year Synchron unveiled "Chiral," a foundation model trained on de-identified neural recordings, with Nvidia's edge-AI platform handling on-device signal processing.
The through-line is clear: the implant is becoming a pipe, and the intelligence that makes it useful is moving into models. That is also why Synchron closed a US$200 million Series D in late 2025 — the company is betting that the bottleneck is data and decoding, not the surgery.
Honest limitations
This piece is based on Synchron's own announcements and company-sourced reporting; it is not a clinical evaluation. The hands-free chat feature has been shown with small numbers of users in an early feasibility study, not a large trial, and broad availability still depends on future FDA review of the full implant system. Performance claims like "conversational speed" are the company's, measured in demonstration settings rather than published peer-reviewed benchmarks. I have not independently verified the end-user error rate of the AI suggestions, and the long-term stability of a vessel-mounted electrode is still being studied. None of this changes the direction of travel, but readers should treat "available to patients everywhere" as years away, not months.
What readers can do now
- If you work in accessibility or care for someone with ALS (肌萎缩侧索硬化) or locked-in syndrome, track Synchron's COMMAND trial updates and FDA filings rather than crowdfunding ads — the real signal is in clinical registries.
- Separate the "implant" story from the "model" story: the durable advantage here is software that predicts intent, which means the relevant competition is OpenAI, Apple, and Nvidia, not just implant makers.
- Watch the data-governance fine print. Synchron says brain data stays off OpenAI's servers — ask the same question of any brain-computer interface (脑机接口) product before a loved one enrolls.