Our Take
A BCI achieving sustained independent use at scale remains rare; Harrell's three-year track record and job performance matter more than the device itself.
Why it matters
Brain-computer interfaces have struggled to move beyond proof-of-concept. Harrell's case shows what sustained real-world use looks like and what features actually matter for daily life.
Do this week
BCI teams: document user workflows and failure modes across the first 1,000+ hours of deployment, not just initial accuracy metrics.
Three years of steady use
Casey Harrell, a person with ALS, has had electrodes implanted in his brain for nearly three years. He first used the brain-computer interface to "speak" in 2023 and has since accumulated thousands of hours of device use. Unlike most BCI deployments, which remain tethered to lab conditions, Harrell now operates the system largely independently, uses it to browse the web, and performs his job. His development team has added new features over time and plans further enhancements.
The team calls Harrell "the first power user of a speech BCI." That label signals a threshold: not a one-off demo, but sustained operation across multiple contexts and without constant technical supervision.
The durability question
Most brain implants demonstrated publicly show capability at a moment in time: a person types a message, controls a cursor, or speaks a word. The implant works. Then the story ends. What rarely appears is the 12-month, three-year, five-year follow-up: does the device still work? Do the electrodes degrade? Does the person keep using it, or does it sit unused?
Harrell's case breaks that pattern. Three years of continuous use, feature expansion, and independent operation suggest the device solves a real problem persistently enough that a user integrates it into daily life. That is different from proving it can work in principle. His statement to MIT Technology Review—"Living with a disease like ALS, you are supposed to have diminished dreams. I do not"—frames the outcome, but the engineering detail is the thousands of hours logged.
The open question is generalizability. Harrell appears to be an outlier: highly engaged, technically capable (or well-supported), and fortunate in his implant's stability. Most ALS patients do not have access to this technology or to the research infrastructure behind it. The story says nothing about failure rates, dropout reasons, or cost. It is a success case, not a field assessment.
Build for persistence, not demos
If you are working on BCI or any implantable neurotechnology, Harrell's trajectory suggests what matters in the second and third year: feature parity (he surfs the web), reliability under real-world conditions, and the ability to add capability without re-surgery. Labs tend to optimize for accuracy on a single task. The user optimizes for independence and scope. Close that gap early, or the device will be abandoned despite working.
Document what drove feature requests, which ones stuck, and why. That knowledge is worth more than a press release about accuracy gains.