logoalt Hacker News

PaulDavisThe1st01/18/20250 repliesview on HN

That's like saying that if an LLM can function by being trailed on 10B words, it can work by being trained on 10k words.