Hey HN, I've built a site to help discover AI research papers by categorising them into research topics which can be explored.

Keeping up with the latest research is hard, particularly if you are interested in more niche research that doesn't make the front page of HN or trend on twitter, so I built this so I can explore the topics I am interested in. There has been a bunch of tools based on arXiv shown on HN before, and of course well known ones like arxiv-sanity, but none that really worked in the topic/timeline based layout that I wanted (I found after starting mine that Semantic Scholar’s topic beta is closest!).

The papers are from arXiv categories most related to AI/ML and goes back to a few years worth of papers, currently updated once daily.

In the backend it uses embedding models to cluster keyphrases from each paper, that is then used to generate the topics. The generated names of the topics is not completely perfect yet, I have improvements to make but have been focused on building the frontend to get something out there!

Right now you can sign in and save papers - in the future I want to add in the ability to follow topics to get personalised updates, along with a bunch of other features.

I am interested to hear if anyone finds the site useful!