Meta has released a second version of Brain2Qwerty, an experimental system that aims to translate brain activity into typed text — without any surgery.
According to MarkTechPost, Brain2Qwerty v2 is a non-invasive brain-to-text pipeline that decodes typed sentences and reaches 61% word accuracy. MarkTechPost reports that Meta has also released open training code alongside the model.
The key word is "non-invasive." Rather than implanting electrodes in the brain — the approach many high-profile brain-computer interfaces rely on — Brain2Qwerty uses magnetoencephalography, or MEG, a scanning technique that measures changes in brain activity from outside the body. As Tom's Hardware explains, the goal is an interface that lets a person control a computer with their mind without needing invasive surgery.
Tom's Hardware describes the new version as showing "promising improvements" over the first, and says the advances could eventually lead to clinical trials. The system is framed as a potential aid for patients — people who might otherwise struggle to communicate or operate a computer.
A few caveats are worth keeping in mind. A 61% word accuracy rate means the system still gets a large share of words wrong, so this is research, not a finished product. MEG scanners are also large, expensive machines, not something a person can wear around — a practical limit the sources note in describing the technology as a scanner rather than a consumer device.
Why it matters: most attention-grabbing brain-computer interfaces require risky brain surgery, so a credible non-invasive path — even an early, imperfect one — could widen who might one day benefit from typing by thought alone.