Dev Log: Cursor-Inspired Arxiv Research Tool

ArXiv (pronounced “archive”) is probably the biggest source of research journals out there. Reading them wasn’t easy for me, it has been 4 years since I last touched linear algebra. I want to dedicate a couple of hours of my week to force myself to be more literate in the industry.

However, I find myself going back and forth between ArXiv and chatGPT just trying to understand what specifically the writer is talking about. I wanted the experience to be similar to Cursor where I can select the parts that I don’t really understand, and have chatGPT explain it to me.

I made this tool. It doesn’t really have a name, the placeholder name atm is Synthetic Trail.

Here’s how it works. Open the html version of the arxiv that you’re looking to learn from. for example: https://arxiv.org/html/2408.13227v1.

Then all you need to do is append r.apurn.com/ behind that. Like this: r.apurn.com/https://arxiv.org/html/2408.13227v1.

Something like this should show up

Then what you can do is select the text on the right side, then press cmd + shift + L to add those as a context to the chat.

I’m working on these things next

• Saving the chat locally based on the arxiv id, so you can go back to your previous messages!

• (maybe) Google OAuth.

• (maybe) Annotation. I want to figure out how I can annotate the text in the page.