AI that recognizes wild monkey faces and serves them cognitive tests on jungle touchscreens sounds like a nature-doc B-plot. It's actually a proof-of-concept for a loop that's about to reshape how we deliver neuro-rehab at home. CapuchinAI (Emory University + Georgia Tech, American Journal of Primatology) automates cognitive testing for capuchins in situ: computer vision identifies the individual, a touchscreen presents an adaptive task, the monkey responds, the model updates. No lab, no handlers, no dropout. The same loop — identify → stimulate → measure → adapt — is the engine behind AmblyoPunch, a Meta Quest VR game for amblyopia (lazy eye). Instead of monkeys, it's you. Instead of fruit-reward tasks, it's punching Gabor-pattern coins (visible only to the weaker eye) and dodging red spikes (requiring both eyes). Four stages. Fifteen minutes a day. No patch. The jungle lab proved the architecture; the living room applies it. https://www.meta.com/en-gb/experiences/amblyopunch/1239507485902689/