Meta’s brain-to-text work and Neuralink’s latest implant-surgery update are a long way off everyday headset control. But both have made me think about what XR may need for those scenarios where hands, eyes, touchpads, and even our voices start getting in the way.
I keep coming back to XR input because so much else in the market seems to be moving faster than the way people actually control these devices. Headsets get lighter, displays improve, AI and XR are increasingly intertwined – and users still need an efficient way to select, type, move, approve, dismiss and navigate.
Meta and Neuralink aren’t building XR controllers just yet, but both companies are working on ways to turn brain activity into digital action, and with the inherent limits of voice, eye tracking, hand tracking, controllers and touchpads – why shouldn’t our thoughts themselves be the future controllers of XR tech?
Meta Is Working on Brain-to-Text
Meta’s Brain2Qwerty v2 is about non-invasive communication.
The MEG scanner system uses AI to decode brain activity into text from non-invasive brain recordings. Meta says it trained Brain2Qwerty v2 on around 22,000 sentences from nine volunteers, with each participant recorded for 10 hours while wearing a magnetoencephalography, or MEG, device and typing. The system reached 61% word accuracy overall, rising to 78% for the best participant.
For XR, the temptation is obvious. A spatial computer that can respond to intended words or commands would be a different kind of access tool, especially for people who can’t rely on hands, speech, gaze or handheld controls.
The reality is still very far away from that. MEG remains specialist brain-recording equipment, and Meta is framing Brain2Qwerty as communication research, especially for people with brain lesions that prevent them from communicating.
Even so, it’s fun to imagine. A lot of XR input still assumes users can speak clearly, move their hands, hold a controller, look steadily at a target or reach for a phone, but there are enteprise use cases out there where these will be difficult or impossible.
Brain-to-text research sits outside normal headset design today, but it touches on a future version of the same access problem.
Musk’s Neuralink Is Taking the Implant Route
The Neuralink tech is even more sci-fi.
University Health Network says Neuralink’s CAN-PRIME study is evaluating the safety and functionality of the company’s implant and surgical robot, with the aim of helping people with quadriplegia control external devices using their thoughts. UHN also says trial participants are using Neuralink’s implant to control computers, smartphones and robotic arms in daily life.
The latest Neuralink development focuses on the surgical implantation of the device. The company says it performed its first transdural (i.e. passing through the brain’s outer membrane) Neuralink surgery in May 2026 alongside Dr. Andres Lozano at UHN in Toronto.
Obviously, surgical brain implants aren’t practical for regular users of XR tech, but is there a future where the receiver is external? Perhaps attached to the temple or another part of the head, or integrated onto the arm of your smart glasses? It doesn’t seem beyond the realm of possibility.
The take-home message is that Neuralink is trying to turn brain activity into computer action. Today, that belongs in medical and assistive technology, but in the future, could it sit near spatial interfaces in environments where the traditonal methods of XR device control present real operational challenges?
Far Away, But Technically Plausible
Meta and Neuralink are working from very different starting points. Meta is exploring non-invasive brain-to-text. Neuralink is testing implanted control for medical applications.
Both are still early, specialist technologies, and neither is close to normal XR use. But the idea is no longer pure sci-fi. If brain activity can be decoded into text, commands or device actions, it is reasonable to imagine those signals eventually controlling spatial interfaces as well.
The hard parts are probably safety, accuracy, latency, hardware, privacy, regulation and whether anyone outside medical or accessibility use cases would accept the trade-offs. The basic control idea still makes sense: XR needs ways for people to act on digital content without always relying on hands, voice, gaze, controllers or touchpads.
Brain-based input may stay niche for a long time, but I can imagine a world where it could still become part of the future XR control mix, especially for users and environments where normal input methods don’t fit.
