Our Take
Three years of unsupervised home use proves BCIs can survive long enough to matter for daily life—but Harrell remains a single data point, and no one knows if his implant will keep working.
Why it matters
Brain-computer interfaces have been stuck in the lab because no one had evidence they'd last or stay useful without constant researcher intervention. Harrell's sustained independence is the first real-world proof of durability and usability that matters to patients weighing invasive surgery.
Do this week
BCI researchers: Document failure modes and longevity data from every long-term implant, not just wins—the field needs actuarial tables before patient recruiting can be ethical.
Three years of unsupervised BCI use, 3,800 hours and counting
Casey Harrell, a 48-year-old environmental activist with ALS, received four electrode arrays (256 electrodes total) implanted in his speech motor cortex in July 2023. Within one month, the UC Davis team decoded his brain activity into a 50-word vocabulary with 99.6% accuracy. That vocabulary expanded to 125,000 words at 97.5% accuracy.
The critical milestone: in the first 22.6 months after implantation, Harrell used the device for more than 3,800 hours at home without any researchers present (per UC Davis reporting in Nature Medicine). His care partner can now independently connect and disconnect him from the system each morning. He uses it to send emails, surf the web, perform his job, and read to his seven-year-old daughter.
The team reported 99% accuracy on the latest iteration of the decoder. Harrell requested and received features like a "privacy mode" that auto-deletes decoded text and a profanity filter for conversations with his daughter. The UC Davis team is working toward a "brain-to-voice" system that would generate natural speech complete with intonation and emotional tone.
Long-term durability changes the BCI calculus for patients
BCIs have been proven in controlled settings for years. What has been missing is evidence that they work reliably for months or years without a research team on-site. Scar tissue around electrodes, brain degeneration, and signal drift are known failure modes. Harrell's sustained independent use suggests that at least in some cases, the implant can remain stable and functional beyond the typical clinical trial window.
Mariska Vansteesel, a BCI researcher at Utrecht Medical Center not involved in the trial, flagged the real barrier: "For these technologies to be relevant for patients, we really need to test them in settings in which they will eventually be used." Harrell's home use without continuous researcher presence clears that bar in one patient.
But Vansteesel also notes caution: a prior patient in her own work used a fully implanted BCI for seven years before signal degradation made it unusable, apparently due to brain degeneration. Harrell's success does not guarantee the same outcome for others. Jane Huggins, developing noninvasive BCIs at the University of Michigan, points out that many people with progressive conditions like ALS actively avoid hospital stays and invasive surgery, regardless of the technology's efficacy.
The data gap that still matters
Harrell's case is a proof-of-concept for sustained independent use, not a guarantor of safety or reproducibility. The field now has one well-documented patient with 3,800+ hours of home use. That is not enough to predict failure rates, optimal implant locations, patient selection criteria, or long-term signal stability across a population.
Any research team recruiting the next cohort of BCI patients should treat Harrell's data as a lower bound on feasibility, not an upper bound on risk. The ethical bar for informed consent requires actuarial data: What fraction of implants fail by year two? Year five? What degenerative triggers are visible early? Which patients are likely to achieve independence versus those who plateau? Until those numbers exist, individual successes remain stories, not predictions.