Meta can now decode typed sentences from brain signals without cutting anyone open.
The company unveiled Brain2Qwerty v2 on Monday, a non-invasive system that reads the electrical signals a person's brain produces while typing and converts them into text. Unlike implanted brain-computer interfaces — the kind Neuralink requires — this approach works from the outside. Meta has been developing Brain2Qwerty for at least one prior generation, and the second version marks a meaningful step toward practical, surgery-free brain-to-text input.
The audience Meta has in mind is people who have lost the ability to type — through ALS, paralysis, or similar conditions. That makes the system's core limitation more than a technical footnote: Brain2Qwerty learns from typing sessions, which means it needs a user who can type in order to calibrate. The people who would benefit most are precisely those who cannot provide that training data.
Meta is not alone in this space — Neuralink, Synchron, and academic labs have spent years on brain-computer interfaces, mostly with implants. A non-invasive approach that actually works would leapfrog all of them on accessibility. But "works" is doing a lot of lifting here: decoding intent from outside the skull is a harder signal problem than reading from electrodes placed directly on the brain, and the training bottleneck suggests the system is not yet ready for the people it is meant to serve.