Turning Phone Data Into Brain Training

video1.0<iframe src="https://www.loom.com/embed/566cdf3340ef40b0afff80877da728a2" frameborder="0" width="1670" height="1252" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12521670Loomhttps://www.loom.com12521670https://cdn.loom.com/os/loom-media-content/loom/20e455cd-737a-45b4-bada-6241d911b5c7/f9b61a90-04be-4120-90c1-cc90bb51b5ca/sessions-thumbnails-566cdf3340ef40b0afff80877da728a2-e41a3d8ce631eef8.gif/binary?authToken=eyJraWQiOiJtaWNyb3MtZ3JvdXAvdGRwLW9zLWFyY2hldHlwZS84ZG5xamZxNGNrNDRsMW5oIiwiYWxnIjoiUlMyNTYifQ.eyJicyI6Imxvb20tbWVkaWEtY29udGVudCIsImF1ZCI6InRkcC1vcy1hcmNoZXR5cGUiLCJuYmYiOjE3OTM5OTY4NDEsInRpIjoiZDY0NjJkMmYtMWI5ZS00N2M1LWJiZmQtMTBkNzYyYzExNDQ2IiwiYWRpIjoibG9vbS9sb29tIiwiaXNzIjoibWljcm9zLWdyb3VwL3RkcC1vcy1hcmNoZXR5cGUiLCJwaWQiOiJiNTdjYzZkYy00Mzk3LTQ1NzctYmY4OC00MTgxNTVlMGZiMWMiLCJ3YSI6ImZhbHNlIiwib2lkIjoic2Vzc2lvbnMtdGh1bWJuYWlscy01NjZjZGYzMzQwZWY0MGIwYWZmZjgwODc3ZGE3MjhhMi1lNDFhM2Q4Y2U2MzFlZWY4LmdpZiIsImV4cCI6MTc5Mzk5NzQ0MSwiaWF0IjoxNzkzOTk2ODQxLCJqdGkiOiJiZmU2NGVhNC02NTJhLTRlMzItYTU3YS1iOGExNjcxNTQ5NDgifQ.HUIqhRMVgQBipUJkznfm3VDwqIe06A-WI8n9dUAUpAnO6VFTT4681XgzUDzQtoACmgXUz7VCaFIfv5vVHrfZphWIieM2VC2f2k82KyWQRqsIzbgD6pzy9JUyKbW5AeBqW709Tqj4D3U_j0iHo96IlLz-BrXXPFIB7ZxKx5Rq3TW32mtzN1N7Emsc1Yd2neb9EyOOnPmp9o9Pks5g12dqe5o-steAlWJqd5TqjbzSdhlBcZWdaTzPW0O5J8mSlSPeeyMPCuKHrI9M2nPalB6p9kGiqOAxQCFmmo-8mb_LgiO4iFgoGp3GePuv0unBrujpCnCB7Fw-_TOY7FAyb9zKpA&Expires=1793997442&Signature=FAIKYOGDabZzs6ajKcfKDcF%7EJImepo7wHAFnQfU81TDXSQwV5NNRUt0vxbU84%7EqyvxRbaWnQLtxiPeKOPz5jDJFT1Nh-RVofglXWGCQi6dQ3gOKkRX2CGfqpaginABqCyw7rUSe7MpXYaHvdTxs8CEYn5XXFugpnzybJ-gsUZ9TQSgQjAyczGK07Gds2Tc2UUjkpSOGBtU3L%7E5ghX-mj70MUCs6hWqvIrskngcgnSx%7EWUkXuvq%7E6pyPik1cvRmRV-UzpaJGPmF0%7EYpyrtmj4FBY3KbnwDV9qi%7E3axc9QKwPyETyA2G9OOG-z3E%7ENNSmlqq4yP0ZVfLel1MRtQ7YkXw__&Key-Pair-Id=KQOSYIR44AIC0121.607Turning Phone Data Into Brain TrainingThis Loom demonstrates using a brain encoder model, TRIBE V2, to predict brain activity from live video and audio captured on a phone. The system runs on three H100 workers on Modal to optimize latency, producing predicted brain activity delayed by about 15 seconds. The author then clicks a button to have Gemini generate a game designed to activate brain areas that have been less activated, with the idea that scaling to smart glasses could help prevent conditions linked to brain atrophy such as Alzheimer’s. The generator takes about 45 seconds end to end to create the game.