This is so true! Even a few years ago, these benchmarks would have been against pandas (instead of polaes and duckdb) and would likely have looked very different.