Basement AI lab captures 10,000 hours of brain scans to train thought-to-text AI models — largest known neural dataset collected from thousands of humans over six months

A San Francisco start-up, Conduit, has gathered a massive neural dataset of 10,000 hours from thousands of individuals over six months in a basement lab. The data is being used to train thought-to-text AI models that decode brain activity into text. Participants engage in two-hour sessions, conversing with AI through speech or typing. Conduit designed custom hardware combining EEG, fNIRS, and other sensors to optimize signal coverage. As the dataset grew, the model began generalizing across different conditions, reducing the need for aggressive noise reduction. Operating costs decreased as the process scaled, with Conduit focusing on model training and planning to reveal its decoding system in the future.