you are correct, its supposed to generate code for things that are discussed in the paper. the app starts with an example paper, but you can give it any arxiv link. it works best when the task you tell it to implement is actually discussed in the paper, or else it will hallucinate random things. its no where near good as of now but we are constantly improving it

I tried. It didn’t generate any code. It felt as though I got a reply back from ChatGPT.

hmm that's weird. did you make sure that you typed "implement: <task>" ? it has to be in that exact format for it to generate code. we're gonna make this experience much better soon