A study published in Nature Neuroscience on March 17th reports that a team from Massachusetts General Hospital has developed a new brain-computer interface device capable of translating in real-time the neural activity associated with 'attempting to type' in the brains of paralyzed patients into text. The device was implanted into the motor cortex of two quadriplegic patients, where electrodes recorded brain activity and a deep neural network model was used to predict input characters. Experiments showed that one participant could output 110 characters (approximately 22 words) per minute, reaching 81% of the typing speed of able-bodied individuals on smartphones, with an error rate of only 1.6%; the other participant could output 47 characters per minute. The system requires training with only about 30 sentences to operate effectively, offering hope for providing paralyzed patients with a more natural and efficient means of communication.