Researchers have unveiled a brain-computer interface that can decode a person’s unspoken words with up to 74% accuracy, marking the first time a device has translated so-called inner speech in real time. The system enabled a woman who lost the ability to talk after a stroke 18 years ago to generate verbal output by simply thinking of the words, offering a potential communication lifeline for patients with severe movement disorders. The interface works by isolating neural signals associated with speech planning and routing them through a decoding algorithm that converts those signals into text or synthesized voice without any muscular movement. Scientists say the approach could eventually give people with paralysis or locked-in syndrome a faster and more natural way to communicate than current eye-tracking or typing methods. Separately, a team in China reported developing a stiffness-tunable "Neurotentacle" probe that reduces tissue damage during implantation by 74%. The flexible design addresses a major obstacle to long-term use of invasive BCIs and underscores the accelerating pace of innovation in neurotechnology.