From the top of a sky scraper you may have a glorious view over the city and landscape, but you can hardly see what happens down in the street. If you look at early mathematics from the height of our time you are bound to miss many a simple but clever method. Artificial Intelligence failed writing a computer program that could make a robot walk. Embodiment is the answer, a bottom up approach complementing the powerful but limited top down approach. In the same way experimental archaeology complements classical archaeology, and I expand this idea to experimental history and archaeology of early mathematics.