Yeah, that makes sense. One fundamental design mistake they made with jupyter notebooks is that it's stored in a JSON with all the outputs in there, so git is getting pretty confused and every run produces changes in the metadata.