Longtime ChatGPT user trying Claude for a big markdown textbook library — hit the project knowledge cap immediately

Been a ChatGPT/Codex user since basically day one, finally giving Claude a real shot, and I ran straight into something I didn't expect. My setup: 8 textbooks (4 biochem, 3 ochem) converted to markdown with figures stripped out. On ChatGPT this was a non-issue — I have GB of storage and the only thing I ever bumped into was the file count per project. Uploaded to a Claude project and got "Project knowledge exceeds maximum" almost right away after only 7 md files. After reading up on it, I get why : Claude loads project knowledge into the context window rather than sticking it in a retrieval index, so the cap is summed tokens across all files, not a file count or a storage quota. RAG kicks in automatically past a threshold and expands it though still can't bypass the upload limit for project context unless individual uploads into the chat, but that is the whole purpose, which is for the ai to find it for me. Options I've found so far: Split into two projects (biochem / ochem). Google Drive connector so files live in the cloud and Claude searches them per query (chatgpt just uses its own cloud I guess) Cowork/desktop project pointed at a local folder — no cap at all since nothing gets preloaded, but then I can't use my phone unless computer is on or plugged in somewhere. Just use NotebookLM, which is arguably built for exactly this Anyone doing something smarter that I'm missing? Also is there a remote like chatgpt has where I can connect my phone and type from there to my computer to where it is running. I see dispatch, but I do not see the conversations.