Ask HN: How do you QA GPT-3 apps?
If your GPT-3 powered app relies on custom crafted prompts, how do you conduct QA on your app to ensure reliability? It seems like you can get different results for the same prompt, so I'm curious if this is a challenge or not.