“Researchers implanted five electrode arrays in the brain of Keith Thomas, a man paralyzed below the chest after a 2020 diving accident, creating a system that allowed him to move his own paralyzed hand via an AI decoder and feel sensation through stimulation of his sensory cortex, while a reinforcement learning system regulated his grip…”

The implant configuration is confirmed by the authors' own methods text. The participant was implanted with five 10x10, 1.5 mm platinum microelectrode arrays with sputtered iridium oxide film coated tips in the hand areas of M1 (two 64-channel arrays, 128 total) and S1 (three sparsely populated arrays, 32 channels each, 96 total) in the left hemisphere. The institution describes the same hardware as five microchips carrying 224 electrodes in total. The AI architecture is as described. The system uses a nested closed-loop control architecture combining a long short-term memory (LSTM) neural network and a deep reinforcement learning (RL) agent, with movement intentions decoded from two arrays in primary motor cortex while stimulation patterns are delivered to up to three arrays in S1. The movement pathway does drive his own muscles. When Thomas thinks about moving his hand, AI translates those neural signals into patterns of electrical stimulation delivered to muscles in his forearm, moving his own paralyzed hand, while force sensors embedded in a custom 3D-printed orthotic device measure pressure during grasping and trigger stimulation in the brain's sensory cortex. The eggshell result is real and quantified. The authors set up a task involving hollow eggshells in which the participant was blinded to the eggshell and his hand but could see a real-time plot of the grasp force; with the deep-RL agent he grasped the eggshell with consistent and safe force levels, and without it the force was highly variable and often above the predefined limit, with 87% success with RL versus 27% without (P = 0.012). Nature Medicine's own summary scopes it: under controlled conditions, the participant demonstrated iBCI-mediated high-precision grasping, successfully manipulating fragile hollow eggshells in real time. Critically, the fine-grasp force was produced by a powered device, not by his own muscles. The methods state it directly: "To increase grasping forces, we developed a custom 3D-printed low-profile active orthosis (AO) that used an artificial tendon system to flex the index and middle fingers for naturalistic grasping. A force sensor embedded in the AO measured grasp strength, which was conveyed to the participant in real-time through S1 ICMS. Decoded motor intent was used to activate the AO which modulated the grasp aperture." A 14 kg servo motor was mounted on the AO over the flexor digitorum superficialis region. The published figure legend for the eggshell demonstration specifies the effector: "e, Demonstration of precision grasp and lift of a hollow eggshell using the AO." On the strength and sensation persistence findings in the post's caption: over 35 weeks of the intervention Thomas achieved statistically significant increases of 86% in right arm strength and 62% in left arm strength, and a man who could not lift his hands to his face could now independently scratch his nose and wipe his mouth. The measure was specifically isometric elbow flexion force driven by transcutaneous spinal cord stimulation: within 15 weeks of administering tSCS alone during activity-based training there was a significant 70% and 25% increase in volitional force output during isometric right and left elbow flexion respectively. On persistence, the paper and the press release say different things. The institution states that recovery persisted for months after stimulation ended, suggesting rewiring of damaged neural circuits. Reporting on the paper's sensory results is narrower: the improvement persisted for more than two months after stimulation ended, although he sometimes felt pressure without correctly locating it, and sensation did not return in the tested thumb or finger areas. The two-year figure appears as an investigator remark in the press release, not as a reported study result: "Remarkably, in a recent follow-up, it was found that these gains were still present after more than two years. This is incredibly encouraging," said Dr. Bouton. Participant and injury details check out. The trial involved a 42-year-old man with complete tetraplegia following a diving accident, with injury at the C4 sensory and C5 motor levels, leaving him unable to lift his hands to his face, grasp objects or feel his hands and lower forearms. He broke his neck in a 2020 diving accident and enrolled in the three-year trial 13 months later. The authors state the study's own limit: "This evaluation of the DNB system was limited to one participant."