In research presented by Peter Janssen’s team at KU Leuven, three rhesus macaque monkeys each received three Utah arrays—96 electrodes per array—targeting not only the primary motor cortex but also the dorsal and ventral premotor cortices. This methodology, which records from both movement execution and higher-order planning regions, allowed an AI decoder to translate raw neural signals into smooth navigation commands inside multiple virtual environments. The monkeys controlled a sphere from a first-person viewpoint, piloted animated avatars from a third-person game-like perspective, and progressed to opening doors and traversing virtual buildings. Sample size was small (n=3), all animals were healthy rather than paralysis models, and the work appears to be conference data rather than a fully peer-reviewed publication as of late 2024.

This goes well beyond earlier primate BCI demonstrations. Classic experiments, such as Miguel Nicolelis’s 2000s Duke University work (Science, 2008) where monkeys controlled robotic arms via motor cortex signals, still required the brain to simulate physical execution. Similarly, the BrainGate consortium’s peer-reviewed human trials (Nature, 2021) relied on imagined hand or finger movements, producing what study participants often described as an exhausting, non-intuitive mental gymnastics routine that took weeks to master.

The New Scientist article accurately reports the premotor targeting but underplays two critical patterns. First, tapping abstract planning signals creates genuine context invariance: the same neural ensemble can drive a first-person sphere, a third-person monkey avatar, or a future wheelchair without retraining. Second, it directly addresses the “ear-wiggling” frustration repeatedly documented in Neuralink’s early human coverage (Musk’s 2024 announcements and subsequent New York Times reporting). By decoding intent rather than execution, cognitive load drops, potentially shortening the months-long calibration period that has plagued invasive BCIs.

Limitations remain stark. Electrode arrays provoke gliosis; signal quality commonly degrades within 12–24 months, a problem documented across nearly every long-term primate and human implant study to date. Human premotor mapping is less precise than in macaques, exactly as Janssen noted. Ethical questions around elective neurosurgery for non-life-threatening applications also deserve deeper scrutiny than popular coverage has offered.

Synthesizing this with Andrew Jackson’s commentary at Newcastle University and the 2021 BrainGate Lancet paper on tetraplegic participants, a clearer picture emerges: the field is shifting from “can you move a cursor?” to “can you fluidly inhabit digital and physical spaces the way an able-bodied person does?” This monkey work marks a genuine inflection point toward that goal. If premotor targeting proves stable in humans, the pathway to intuitive prosthetic limbs, thought-controlled wheelchairs, and even immersive virtual worlds for people with severe mobility loss accelerates dramatically. The leap is not that monkeys can play a video game with their minds—it is that they do so with the same effortless flexibility our own brains use when we simply decide to walk across a room.