Feed it right | Master AI Automation in 4 hours Master AI Automation in 4 hours Course About Ayush Modules Sample chapter Toolbox The Microcap Minute Module 08: Documents, Data & Research / Chapter 1 Feed it right Watch first, then read.
Same lesson, your pace.
What you will learn - The three doors a document takes into a model, and when each wins - NotebookLM: the grounded reading assistant - Context hygiene for big piles Three doors for one document Door 1, Paste.
Under ~20 pages of mostly text, pasting is fastest.
It survives the trip well, loses tables/layout, and (Module 11 preview) should never carry sensitive personal data.
Door 2, Upload to chat.
Claude, ChatGPT and Kimi all accept PDFs directly; layout, tables and images survive.
Best for one-off Q&A on big files.
Files persist in that chat, or pin them to a Project (Module 2 Chapter 5) so every conversation there inherits them.
Google's free research assistant built entirely around your sources: upload many documents and it answers only from them , citing exact passages.
Hallucination collapses because the desk holds your papers and little else, the strongest anti-hallucination architecture in Module 1's terms.
Bonus: it generates study guides, FAQs and even podcast-style audio overviews of your material.
Decision rule: quick question on short text → paste; heavy file or recurring project material → upload/Project; serious multi-source study → NotebookLM.
Context hygiene Big piles degrade answers (attention dilutes across a crowded desk): Split by purpose , "chapter 3 only" beats "entire textbook" for specific questions Say what to ignore , "use only section 4.2" focuses the model Name your files meaningfully , 2026-physics-syllabus.pdf beats doc_final_v3.pdf ; models read filenames too Re-anchor long chats , after many exchanges, restate key constraints; early instructions fade Try it yourself Take any real PDF (syllabus, circular, chapter).