A speech brain computer interface, or BCI, reads brain signals and turns them into synthesized words. Casey Harrell has used one for almost three years, longer than any patient before him, and has begun to reshape the technology he depends on to communicate.
Casey Harrell has had electrodes implanted in his brain for almost three years. He now uses the speech brain-computer interface, or BCI, largely on his own: to work at his job, to browse the web, to push the research team for new features. The team that built the device has a name for what Harrell has become. They call him their first "power user," a term borrowed from software, not medicine, that captures something the field has rarely seen: a patient with amyotrophic lateral sclerosis (ALS) who has logged so many hours on an experimental speech implant that he is now co-authoring the technology, not just receiving it.
ALS is a progressive neurodegenerative disease that strips away voluntary muscle control, including the muscles that make speech possible. The BCI works around that loss by reading electrical activity from the brain's speech-motor areas and translating it, via machine-learning software, into synthesized words and on-screen actions. Harrell first used the system to "speak" in 2023, according to MIT Technology Review's Download newsletter. Since then he has piled up thousands of hours of use, the bulk of it without hands-on help from the researchers.
The shift in Harrell's role, from research subject to something closer to a product manager, is the part of the story that doesn't show up in the device's spec sheet. He has driven changes the team hadn't planned: smoother word prediction, more reliable web navigation, a workflow that no longer requires a technician in the room. "Living with a disease like ALS, you are supposed to have diminished dreams. I do not," Harrell told MIT Technology Review. The line lands because the agency it describes is operational, not aspirational. He is using the device to do the things he was told the disease would take from him.
The "power user" label, though, is the team's vocabulary, not Harrell's self-description, and the framing it invites deserves pressure. Harrell is, by any reasonable reading, the only person on earth with this much logged time on this particular speech BCI. That makes the case a study in what is possible, and a warning about what is not yet generalizable. The device remains experimental. Latency, the gap between thought and synthesized word, is still measured in perceptible milliseconds. The vocabulary the system can reliably decode is smaller than a typical adult's working lexicon. The electrodes are surgical, and their long-term stability inside living brain tissue is itself an open engineering question. No independent peer-reviewed data set yet matches the depth of Harrell's hours.
None of that erases what Harrell has done with the time he has been given on the device. The thousands of hours have produced something the field's usual clinical-trial frame cannot easily capture: a patient who has outlasted the assumptions baked into the protocol, and in doing so, has started to define the next questions. The research team continues to add features. Harrell continues to ask for them.
What is emerging, slowly, is a different model for how an assistive device gets built. Not a clinic hands a finished product to a patient, and not a patient endures a clinical trial until the data is in. Something in between: a single user with a degenerative disease, logging the kind of hours that turn a prototype into something a person can actually live inside. The next test of that model will be whether a second patient can do what Harrell has done, and whether a third can push the device further still.