The poster raises a specific question: where can one find focused, open-source, properly filtered deep learning corpora? The scope of the need includes mathematics, human language (including historically incomplete languages), physics, and all scientific categories, with the key words being "dense, educational, and clean".