[arpowers]

Could use some examples of the improved output. It shouldnt be that hard.

What are each models strengths? I know GPT4 has a strength with code.

[simonw]

I linked to an example of Claude 3 beating GPT-4 on code from the article.

GPT-4: https://chat.openai.com/share/117fb1ad-6361-41e2-be59-110f32...

Claude 3 Opus: https://gist.github.com/simonw/2002e2b56a97053bd9302a34e0b83...

The GPT-4 code didn't work - it was missing some async keywords.

The Claude 3 Opus code worked perfectly.