Accessibility looks different when it starts inside the skull, not on a laptop screen. In a quiet lab, an implanted brain-computer interface feeds streams of neural spikes into a new decoding system, and for one person with advanced ALS, those signals now stand in for a lost voice during a full-time remote job.
The striking point is not the metal in the head but the math around it. UC Davis researchers kept the same intracortical electrodes and focused instead on a machine learning pipeline that links motor cortex activity to words, using probabilistic language models and neural signal processing to infer intended sentences with about 92 percent accuracy. Short pauses and error checks replace small talk, yet the system runs fast enough for live workplace chats, email replies and document edits through a standard computer interface.
What looks like ordinary knowledge work on a videoconference window is, under the hood, a continuous loop between cortex and code. Decoders trained on repeated imagined speech and attempted hand movements perform feature extraction on firing rates, then feed into classifiers that select letters and words, which a text-to-speech layer or on-screen cursor sends on to colleagues. The hardware may be familiar to neuroprosthetics teams, but in combining stable decoding with practical communication speed, this setup pushes the idea of reasonable workplace accommodation into territory that once belonged only to speculative fiction.
