Give Me 13 Minutes. I'll Teach You 80% of NotebookLM
Video by Linking Your Thinking with Nick Milo on YouTube · 13 min · uploaded February 2026
Notes and analysis by BrainTube, written from the video. Not affiliated with Linking Your Thinking with Nick Milo. Watch on YouTube
In short
Nick Milo walks through NotebookLM's three panels (sources, chat, studio), eight practical tips and two common mistakes, then shows how he uses it as a short-lived research layer in front of his Obsidian vault. The tour is accurate and the advice to keep notebooks narrow and to treat NotebookLM as a literal researcher is sound. He overstates how far citations remove hallucinations, and he skips the cost of notebooks being separate silos.
The argument
Milo's argument is that NotebookLM is best used for orientation, not output. You load a focused set of sources, ask questions whose answers are grounded in those sources with clickable citations, and get the shape of a topic quickly. The studio's podcasts, videos, slide decks and quizzes are fun but of limited value, because turning knowledge into your own understanding happens when you rewrite and connect it yourself. So he treats each notebook as temporary. When he has the gist, he retires it and moves his best insights by hand into Obsidian, filing them as long-term knowledge, active project material or a dated log entry. The manual step is the point: it forces him to do the connecting.
Key ideas, with timestamps
- 0:57
Three panels, one loop
Sources on the left, chat in the middle, studio on the right. He maps these to his ARC loop of add, relate and communicate. Sources can be PDFs, documents, YouTube URLs, websites, audio files and Google Drive files, and the chat answers only from what you added.
- 1:27
Limits and privacy
As of the February 2026 upload, he cites up to 100 notebooks, 50 sources per notebook and up to 25 million words combined. On privacy, he says Google promises that sources and chats are not used for training, and he rates it as roughly as safe as Google Docs or Gmail. Limits vary by plan and change over time, so check the current numbers.
- 2:51
Citations are the reason to use it
Every answer carries numbered citations that jump to the passage it drew on. He contrasts this with general chatbots whose confident answers are hard to verify. Clicking through lets you check whether the model misread or overgeneralized a source.
- 3:46
Save, toggle, paste, rename
Tips one to four: save a good answer as a note and convert it to a source, use the checkboxes to limit which sources an answer draws from, paste raw text when a source will not upload, and rename sources so you remember what each one holds when toggling.
- 5:08
The 50-source workaround
When a complex topic hits the source cap, summarize a few sources you have already worked through, save the summary to a note, convert it to a source and delete the originals. Tip six, at 5:38, is to configure the notebook's chat persona, such as a marketing expert. Tips seven and eight cover the mobile app for quick capture and sharing notebooks.
- 7:04
Two mistakes
First, putting too many unrelated sources in one notebook, especially with all of them toggled on, which weakens answers. Split the topic into several notebooks. Second, treating NotebookLM as a creative partner. He describes it as a hyper-literal, diligent researcher that will say it does not know, which is good for accuracy and bad for brainstorming.
- 8:28
The studio is the least important panel
Audio overviews, video overviews, infographics, slide decks, mind maps, reports, flashcards and quizzes. He likes the mind map because clicking a node starts a chat on that subtopic, and flashcards help students. But at 9:53 he says producing outputs is of limited value, because it is not how knowledge becomes your own.
- 10:20
A research layer in front of Obsidian
He opens a notebook for a new area, adds a few videos and articles, and asks what the main idea is, where the sources disagree and what to watch for. Then at 11:44 he retires the notebook and hand-copies insights into Obsidian: evergreen knowledge in his Atlas folder, project material in Efforts, and a dated log in Calendar.
What it gets right
The framing of NotebookLM as a literal researcher is the most useful idea for new users. People open it expecting a brainstorming partner and are disappointed. Knowing that its strength is staying close to the sources sets the right expectation and points you to the right questions: what do these sources say, where do they disagree, what is missing.
The advice on notebook scope is also correct. Retrieval over a focused set of sources gives cleaner answers than retrieval over a pile of loosely related files. Toggling sources off is the cheapest way to sharpen an answer, and most users never touch those checkboxes.
His workflow is honest about where value is created. Generating a slide deck from ten sources feels productive, but it skips the step where you decide what matters. Moving insights by hand into his own notes is slower, and that is the point he is making.
What it glosses over
- Citations reduce errors, they do not remove them. He says hallucination is not the case in NotebookLM. Grounding and citations make errors much easier to catch, but a grounded model can still merge two sources, quote out of context or miss a caveat. His own advice to click through is the safeguard, so follow it.
- The source-merging trick has a cost. Replacing originals with your summary frees slots, but later answers will cite your summary, not the original passage. You lose the traceability that he calls the tool's main strength.
- Notebooks are silos. Each notebook only knows its own sources. If a video you added last month is relevant to today's question, you have to remember which notebook it was in. His habit of retiring notebooks makes this worse: the sources are effectively gone unless you copied the insight.
- YouTube sources depend on captions. A video source is useful only if its transcript can be pulled. The paste-text tip helps when an upload fails, but he does not say why videos sometimes fail.
- Privacy is a judgment, not a guarantee. The no-training promise, as of the upload, applies to the content you add and your chats. Your data still sits on Google's servers. For client or regulated material, check your account type and your organization's policies before uploading.
How this applies to memory across AI tools
Milo's workflow solves the silo problem with manual effort: he copies what matters into Obsidian. That works if you are disciplined, but the sources themselves stay behind in a retired notebook, and none of it is visible to the other assistants you use. If you research in NotebookLM, write in Claude and code in Cursor, each tool sees only what you paste into it.
The alternative is to keep sources in one place that every assistant can query. BrainTube takes that approach. It saves YouTube videos with their transcripts, articles, PDFs and podcasts into one compiled knowledge base and exposes it as an MCP server, so Claude, ChatGPT, Cursor and other MCP clients can search it across topics. His advice still holds either way: keep your own notes for the thinking, and use retrieval for the sources. For other options, see NotebookLM alternatives for YouTube and chat with YouTube videos.
Who should watch it
Watch it if you have never opened NotebookLM or used it only for audio overviews. The tips from 3:46 to 6:35 are the most practical stretch, and the two mistakes at 7:04 are worth the minute. Obsidian users who want to see how a tool like this fits a notes vault should jump to 10:20. Skip it if you already use NotebookLM weekly, since most of it will be familiar, or if you want detail on the studio outputs, which he covers only briefly.
Frequently asked
- How many sources can you add to NotebookLM?
- As of the video's February 2026 upload, Nick Milo cites 50 sources per notebook, up to 25 million words combined, and up to 100 notebooks. Limits differ by plan and Google changes them, so check the current numbers in your account. If you hit the cap, you can summarize sources you have finished with and replace them with the summary.
- Does NotebookLM hallucinate?
- It answers only from the sources you add and cites the passage behind each claim, which makes errors far less common and much easier to catch than in a general chatbot. It can still misread or merge sources. Click the citations for anything you plan to rely on.
- How do you use NotebookLM with Obsidian?
- One approach, shown by Nick Milo, is to use a notebook to get oriented on a new topic, ask it where sources agree and disagree, then retire it. You copy the insights you want to keep into Obsidian by hand, filing them as long-term notes, project notes or a dated log. The manual rewrite is what makes the knowledge your own.
Keep videos like this one searchable
BrainTube saves YouTube videos with their transcripts into one compiled knowledge base that Claude, ChatGPT, Cursor and other MCP clients can search.
Go deeper
More video notes
Are you the creator and want this page changed or removed? Contact us and we will act within 48 hours.