We should train models on math and philosophy before we show them history. Then they can figure out how to square that circle on their own.