30papers is a reading guide for people with little experience reading research papers. I built it from my own experience trying to read AI research. It was hard to know which papers were essential for understanding the current state of the field. And once you found them, they assumed familiarity with prior work that beginners rarely had.
I took a list of 30 papers recommended by Ilya Sutskever and wrote a summary for each. Inspired by Stripe's edition of Poor Charlie's Almanack, I added inline explanations for key terms. I had relied on AI heavily to explain concepts while reading the papers myself, so I also added buttons to open ChatGPT or Claude with a prewritten prompt.
I posted the site on Hacker News hoping it might help a few people in the same position. It reached the top 10 posts on the front page by the end of the day and has since picked up over 641 upvotes. Traffic spiked to more than 65,000 unique visits in 24 hours. It eventually even attracted the attention of AI researchers at Google. Watching visitors from almost every country show up in Cloudflare analytics was the proudest I have been of a project.
The Hacker News thread also drew criticism, which I tried to take on board without letting it overshadow the success the website had had. Commenters flagged the landing-page background as distracting, so I added options to disable motion and the background, reduced background opacity, and improved LaTeX formatting. After those changes, negative comments gradually gave way to people saying how useful they found the site. I still regularly update the site and check the Hacker News listing for feedback. If you have used the site and have suggestions for new features, please get in touch.
