I scrape arxiv once a day and have a small pipeline that tries to identify all AI-related submissions and summarize them. It’s hit-or-miss currently, but a.) the papers it finds are sometimes pretty interesting when I skim them to check the summaries and b.) it’s a great project to focus my AI learning on and I’m improving it all the time. One day I hope to get it good enough that it’s suitable to post to HN!