logoalt Hacker News

p1eskyesterday at 4:42 PM0 repliesview on HN

How can you train a Claude/Gemini scale model if you’re limited to <10% of the training data?