All Embedding Models Learn the Same Thing / hacker news