Brain-computer interfaces have moved from laboratory demonstration to implanted devices in a small number of patients, with results that are genuinely significant for the people involved.
What has been demonstrated
Individuals with severe paralysis controlling a cursor, typing, and operating software directly, at speeds that make communication practical rather than merely possible.
For someone who had lost the ability to interact with a computer at all, that is a substantial restoration of independence. It should not be diminished by scepticism about the surrounding hype.
The constraints that remain
Implantation is neurosurgery, with the risks that implies. Electrodes provoke a tissue response that degrades signal quality over months and years, which limits device lifetime. Each system requires individual calibration and periodic recalibration.
And the participant numbers are small. These are early-stage trials, not treatments, and the distinction matters for anyone reading a headline as availability.
Reading is not writing
Current devices decode neural activity into commands. That is reading. Writing information into the brain – inducing a specific perception or memory – is a fundamentally harder problem, and progress on it is far behind.
Almost every speculative application people find exciting, from downloaded skills to shared experience, requires writing. None of them is close.
The consumer question
Non-invasive headsets that read electrical activity through the skull exist and work poorly. The skull attenuates and blurs the signal severely, which limits them to coarse measures – attention, relaxation, simple binary choices.
The gap between that and an implanted electrode array is physics, not engineering iteration, and it will not close with better software.
What is worth watching
Longevity data. A device that works for a decade changes the risk calculation entirely; one that degrades in two does not. That figure, more than any demonstration of speed, determines whether this becomes a treatment or stays a trial.