Three years of feedback from Casey Harrell has pushed the UC Davis team to ship accuracy upgrades, a privacy mode, and a profanity filter that keeps him from swearing in front of his daughter.
Casey Harrell doesn't just use his brain-computer interface. He has spent nearly three years becoming what UC Davis researchers call "the first power user" of the device, and in that time, his day-to-day feedback has driven concrete engineering changes to the system that decodes his attempted speech, according to MIT Technology Review's feature on the expanding BCI trial landscape.
Harrell, who has ALS, was implanted with the brain-computer interface, or BCI, in July 2023. The device sits over the part of his cortex that controls speech and translates his attempted movements into words on a screen, fast enough that he can hold a conversation, browse the web, and keep working as a climate activist, per the same MIT Technology Review report. Almost three years in, the team has shipped upgrades he asked for: better decoding accuracy, a privacy mode, and what Harrell calls a "profanity filter" that keeps him from swearing in front of his young daughter when the decoder picks the wrong word.
That last feature is the one that tells you this technology has left the lab. A clinical trial that ships a profanity filter is a clinical trial that has met a real child. It is also the kind of fix no bench test would have produced, because no bench test cares whether a father sounds like himself at the breakfast table. Harrell's role, in the language his UC Davis team uses, is closer to an iterative design partner than a subject, and MIT Technology Review frames his case as a leading edge of a wave of BCI work aimed at restoring communication and independence to people with paralysis.
The technology is not a product. It is also not a promise anymore. The interesting question is what "taking off" actually means when the people doing the takeoff are the same people wearing the device. Harrell calls the implant "nothing short of revolutionary" in his own account to MIT Technology Review, citing his maintained income, his reconnection with friends and family, and being able to read to his daughter. That is patient self-report, not a device specification, and it is the right unit of measure for a man who needed a working voice, not a benchmark.
What the trial does not yet show is whether the iterative-feedback model scales. Harrell is one patient at one academic medical center, and no independent trial-enrollment or regulatory figure was attached to this reporting. The MIT Technology Review feature points to a wider BCI trial landscape, but the readouts needed to compare mechanisms, enrollment, and patient counts across those programs were not part of the source basis here. The next test is whether the redesign loop that produced a profanity filter in Sacramento shows up in trials elsewhere, or whether every BCI program has to discover swearing on its own.