UC Davis Neurointerface: 1.96 million words in 19 months — a new standard for home BCI
We are witnessing a historic breakthrough in the field of brain-computer interfaces (BCI). Researchers from the University of California, Davis have presented data on the long-term, fully autonomous home use of a speech neuroimplant by a patient with amyotrophic lateral sclerosis (ALS). Over 19 months, patient Casey Harrell transmitted 183,060 sentences — nearly 2 million words — at an average speed of 56 words per minute. The key difference of this work from previous ones is that it describes not a laboratory demonstration, but real, everyday use of the system without the constant presence of scientists.
Everyday Operation: 3,800+ Hours of Active Use
Harrell used the neurointerface for more than 3,800 hours. The system became his primary tool for communicating with family, friends, colleagues, and doctors, allowing him to send messages, participate in video calls, use the internet, and maintain full professional employment despite complete paralysis. According to the patient's own assessment, 92% of sentences were decoded as at least "mostly correct." In formal tests where Harrell was shown words on a screen, accuracy exceeded 99% with a vocabulary of 125,000 English words, reaching a peak of 99.2%. After the patient's assistants were allowed to independently connect and disconnect the equipment, the average daily interaction time increased from 3.7 to 9.5 hours.
Technology: From Brain Signal to Synthesized Speech
In 2023, Harrell had four microelectrode arrays implanted in the left precentral gyrus — the area of the cerebral cortex responsible for coordinating speech. The system reads signals from 256 cortical electrodes at the moment the patient attempts to speak. An algorithm converts neural activity into phoneme probabilities every 80 milliseconds, after which a language model selects the most likely sequence of words from a vocabulary of ~125,000 units. The text is displayed on the screen in real time, and after the phrase is completed, the system can vocalize it using a synthesized voice tuned to Harrell's voice before the illness. Notably, the patient also used a cursor decoder to control the computer, operating on signals from the same speech motor cortex, which refutes old notions about the need for different brain areas for speech and movement.
From Lab to Home: A New Stage in BCI Development
Previous work by the team demonstrated high accuracy in laboratory sessions (99.6% with a vocabulary of 50 words). The new publication shifts the focus to long-term, unsupervised use in everyday life. Study co-author Sergey Stavisky noted that 3,800 hours of brain activity recording during system use represents, in his data, the largest individual dataset with resolution at the level of individual neurons. However, the authors emphasize that this is still a single clinical case, and it is unknown how applicable the results are to other patients. The system remains experimental, requires a wired connection and daily setup by trained assistants, and is limited to home use due to the size of the equipment.
My comment: This work is a crucial step from theoretical demonstrations to practical, life-essential tools. 3,800 hours of autonomous operation is not just a number; it is proof that BCI can become a full-fledged part of everyday life, not just a laboratory artifact. However, the problem of scalability and miniaturization remains a key challenge. For now, we are only seeing the beginning of the path toward truly universal and accessible neurointerfaces.