Ask HN: Best foundation model for CLM fine-tuning? | Not Hacker News!