stop pasting documents into Claude Code, you're using a $200 brain as a search engine every time yo...

Stop pasting documents into Claude code. Every time you paste A10 page doc into your session, Claude reads the whole entire thing just to find one section it needs. If you do this enough times, this can be a very expensive habit. Here's the split that fixes it. You can use the notebook LMMCP which handles retrieval. You can load your sources once, specs, past projects, client docs, competitor research. Then they just sit there and they never expire. The best thing is that Claude doesn't see them until it needs them, so you're not wasting context window and notebook does the chunking for us. Claude code handles the execution when it needs something. It queries notebook LM via the MCP server and it gets back the exact chunk it needs, which is like 200 tokens instead of 2000. The contact stays clean, the session runs faster, and notebook LM is completely free to use. I'm using this exact workflow for every single project I've been working on, and I'm noticing a very tangible difference in my token consumption.