Meta AI has built a system called Brain2Qwerty that can reconstruct typed words from non-invasive brain recordings using a helmet covered in sensors, detailed in a paper published on February 6, 2025. The system achieved an average character error rate of 32% when using magnetoencephalography (MEG) sensors, compared to a 67% error rate with EEG-based readings, with the best MEG participants hitting a 19% character error rate. Brain2Qwerty uses a three-stage pipeline: a convolutional module analyzing 500-millisecond windows of brain signals, a transformer processing patterns at the sentence level, and a pretrained language model refining output at the character level. The research involved 35 healthy volunteers who performed typing tasks while wearing MEG or EEG headsets. The original paper explicitly stated that the system does not operate in real time, requiring an entire sentence before decoding. The research team, which included collaborators from Paris Sciences et Lettres University and the Basque Center on Cognition, Brain and Language, positioned this work as a proof-of-concept.