# 10 ways LLMs could help reading [Tweet](https://twitter.com/jessmartin/status/1633145555684585473) Here's 10 (not all of these are LLM-only): 1. Prepend every article with an auto-generated tl;dr 3-bullet point summary. 2. "Chat with the article" ask questions about it using chat. 3. Semantic highlighting layer for most salient sentences/paragraphs in the paper. (closest to a generated summary) 4. For every link in the article, generate a 1-sentence summary of that article as well, let me "peek" at it or embed it in the text. 5. Show me my own relevant highlights/notes/comments that are most closely related to the current article. 6. Turn whole article into bullet-point outline format. Toggle between this mode and full article mode. 7. Generate taxonomic descriptions of the article: howto, diatribe, polemical, fanciful, educational, etc. (think Netflix movie descriptions) 8. Rank closeness to things I've already read. Might want to use this to stay in the niche or to branch out. 9. When I go to add a comment for a highlight, auto-generate 3 possible comments for me. Use my previous comments in the prompt. 10. Tell me when to stop reading. When article is basically repeating same points, warn me to stop.