logoalt Hacker News

jaapztoday at 8:54 AM0 repliesview on HN

When the model produces reasonable results from one prompt, you could assume that it will also return reasonable results through the follow up prompts.