And they've already collected over 100,000 samples, iirc ChatGPT was trained on something like 30,000 samples, so the open models should already be positioned to succeed.