A friend recently told me keeping up with the latest AI research papers feels like a one-legged man in a butt-kicking contest, you just can’t keep up. You try reading a few, gather some info but mostly get distracted ( if you’re like me) and most time dont get back to it or have a growing list of unfinished bookmarks that stare at you like an unhinged racoon.
Being a heavy user of cursor these days (and its 10x’d my side projects easily!), it took me a quick 30 minutes to build and deploy a usable web app that I can store as a repository. High level goals were –
- Snag the latest research papers from a site of choice
- Use an LLM to distill those complex papers into easy-to-understand summaries. Hopefully to partially understand some of the alien hieroglyphics in SOTA papers
- This was an area that still needs a ton of work as the summaries were mainly hit of miss. I end up just navigating to the paper and reading the ones that are interesting as the summaries don’t do them any justice. Still a lot more tinkering here to make this usable including switching out the model with Pegasus for better abstract summaries.
- Presents the link summaries in a clean web interface, so I can quickly scan what’s new and interesting i.e. my own CliffsNotes.
- Keep a low maintenance archive of past papers
- Deploy on to a endpoint I can access it to. I considered hosting it on my own ubuntu server but ended up using GCP Cloud Run instead ( as it was too easy to quickly dockerize and deploy
Hosted here :https://x-papers-625329947111.us
Github: https://github.com/vishwanath79/x_papers
More one-legged antics to follow!