## How do I know what to read? I learn about books to read next mostly through my timeline recommendations on X. Don't know if this is a good thing or not. I am very sensitive to books mentioned in an Youtube video as sources as well. An offline stroll in a bookstore or library is also a good way of randomly encountering new books. Recommendations from friends is also greatly appreciated. Sometimes I also ask AI for suggestions, but I found the quality of the results ragged. --- ## How do I get and prepare the books? First I search for and download the book on Anna's Archive, for which I have a yearly donation so I can skip the waitline. For formats I mostly go for epub or real pdf. Real pdf can be directly parsed by the Gemini models, so no processing is needed. Epub, on the other hand will be converted to txt (or md, doesn't really matter) via a local script that uses pandoc. Sometimes, only scanned pdf is available. In this case, either book is extremely precious so I will read by eye any way, or I will use an OCR service. --- ## How does it work to read a book with LLM? [[How I read a book with LLM]] --- ## What to read with AI, and what to read by myself I only read books whose language has irreplacable human touch. The books that have meaning beyond its message -- but in the way its expressed, and unique and irreplacable human experience embodied, etc. For the sake of transmitting information or dry knowledge / insights, I find it much more efficient to use LLM as a tool to make the reading process much more engaging, direct, active, and efficient. The point of reading with AI is not to replace human reading, but to save time for those books that truly deserve the human reading word by word. More broadly, the point of using any AI is not that humans don't need to anything; but to save time for those uniquely human activities by liberating us from chores that could have been done by machines. --- ## What is this "Deep Summary" Thing?