My sister was angry with Claude this week. Her words: "we're fighting."
She'd gone to it to talk through something personal. Instead it kept dragging in pieces of her business, her work, and conversations that had nothing to do with what she wanted to discuss.
It felt intrusive. Worse, it made the answers useless.
Why is it mixing everything together? I'm not asking about my business. I'm trying to talk about my life.
The problem wasn't that Claude had forgotten her context. It had too much of it.
That sounds backwards. We spend so much effort teaching AI to understand us — what we do, what matters, what we're working on, how we like to be spoken to. Good context usually means better answers.
But more context isn't better context.
Sometimes you don't want AI searching the whole filing cabinet. You want it looking at one folder on the desk.
That's when a tool like NotebookLM becomes useful. It's Google's, and you may see it branded as Gemini Notebook — same tool.
Welcome back to Beyond ChatGPT
Most people still use AI and ChatGPT as if they mean the same thing.
ChatGPT (or Claude) is usually the first AI tool we learn. It's flexible, conversational, impressively capable. I use it constantly. But it's one kind of AI tool — not the definition of AI.
This series is about the capabilities sitting just outside the familiar chat window: tools that research, organize, build, summarize, visualize, and turn information into entirely new formats.
Part 2 is about what happens when you need AI to stay inside a defined set of information. Not your whole life. Not every conversation you've ever had. Just the material that matters for the task in front of you.
They solve different problems
ChatGPT is a broad thinking partner. Depending on your settings, it draws on the current conversation, saved memories, previous chats, project instructions, and files you've given it. That breadth is genuinely useful. It connects ideas, notices patterns, and saves you explaining yourself from scratch every time.
It also creates context collisions.
You want to think through a family situation and work keeps walking into the room. You're planning a trip and it blends details from three different vacations.
You can manage some of this inside ChatGPT — separate Projects, a Temporary Chat, editing what it remembers, telling it explicitly what to ignore. But sometimes the cleaner answer is a tool built around a narrower boundary.
A NotebookLM notebook (you might see it branded as Gemini Notebook, from Google) is built for one body of information. Each notebook is independent. You choose the sources. It works from those sources and the conversation inside that notebook. That's it.
ChatGPT is the colleague who knows your history.
NotebookLM is the specialist you hand a project file to.
And you already know how to do this. You don't give your accountant your medical history. You don't brief the contractor on what's happening at work. You've spent thirty years knowing which information belongs in which room and who needs to see it. This is that same instinct, applied to AI.
What NotebookLM actually is
Google's AI research and learning workspace. You create a notebook and give it source material: PDFs, Word documents, websites, Google Docs and Slides, spreadsheets, PowerPoint files, pasted text, images, audio recordings, public YouTube videos. It can also help you find relevant internet sources and decide what to add.
Then you ask questions across it:
- What are the main themes across these documents?
- Where do these two reports disagree?
- What was decided, and what's still open?
- Explain this in plain English.
- Build me a timeline.
- What should I be asking after reading all this?
Answers are grounded in the sources you chose, with citations that take you back to the supporting material.
It still gets things wrong. It's still AI. But the boundaries are clear: you know what it's working from, and you can check the evidence behind every answer. That's a different experience from asking a general chatbot to assemble everything it might know about a topic — or about you.
The chat isn't the point
This is what makes NotebookLM a real Beyond ChatGPT example. The chat is one part of the tool.
It will turn the same body of information into:
- briefing documents and study guides
- FAQs and timelines
- mind maps
- flashcards and quizzes
- audio overviews that sound like a podcast conversation
- video overviews
- slide decks
- infographics
Reading another 80-page document isn't always the best way to understand something. Some of us want the written brief. Some want to listen while walking the dog. Some need to see how the ideas connect. Some want a deck they can use to explain it to someone else.
NotebookLM isn't answering a question. It's reshaping information into a form you can use.
That's one of the real shifts happening right now. We're moving from AI that talks to us toward AI that builds things from the material we give it.
Where you'd actually use it
Anywhere you have a defined pile of information and need to make sense of it.
A presentation you have to build
I recently prepared a fraud-prevention talk for residents at my father's seniors' home.
My problem wasn't a lack of expertise or material. I had too much of both.
I put credible research, consumer guidance, and my own previous materials into a notebook, then used it to simplify the content for a senior audience and organize the key messages. From that one defined source set I built three things: the presentation, a short handout, and an audio overview people could listen to afterward.
That's what narrowing the context buys you. Better outputs, because the tool is working from the right pile.
A complicated decision
A renovation, a move, a retirement option.
Upload the estimates, inspection reports, pension documents, benefit guides, meeting notes, and your own list of priorities. Ask it to compare the options, name the assumptions underneath them, and write you a decision brief.
It won't make the decision. It'll make the information visible.
The medical example — with one firm boundary
I think about how useful this would have been while helping my mother through a serious illness. Appointment recordings, bloodwork, imaging reports, discharge instructions, medication lists, notes scattered across paper, portals, and family group chats.
I used AI to transcribe appointments and write summaries for relatives who couldn't be in the room. That alone helped enormously.
Today I'd build a private notebook and ask:
What changed between these two discharge summaries?
Which questions are still unresolved across these appointment notes?
It isn't a doctor. Use it to organize, compare, and prepare questions — then take the decisions to someone qualified. The same caution applies to legal and financial material. AI can get you to the appointment better prepared. It shouldn't become the appointment.
And think before uploading anything sensitive. Read the privacy terms for the account you're using, strip identifiers you don't need, and don't assume every AI service treats your data the same way.
Focused doesn't mean accurate
Grounding answers in your sources doesn't eliminate risk.
A bad source is still a bad source. An outdated document is still outdated. A missing page still produces an incomplete answer. And AI still misreads things.
The advantage isn't truth. It's a controlled research environment. You can see what you gave it. You can switch sources on and off. You can follow a citation back to the page. You can keep separate topics in separate notebooks instead of letting everything bleed together.
That makes verification easier. Verification still matters.
Try this in ten minutes
Don't start with your most sensitive material. Pick something low-stakes you already understand.
- Open NotebookLM with your Google account (it may redirect you to Gemini Notebook).
- Create a notebook for one clearly defined topic.
- Add three to five sources — a PDF, a webpage, a document, a YouTube video.
- Ask: What are the five most important ideas across these sources?
- Open the citations and check whether the answer holds up.
- Generate one alternative output — an audio overview, a mind map, a briefing document, a slide deck.
You're not testing whether AI can impress you. You're noticing what changes when you give it a boundary.
The bigger lesson
My sister's frustration wasn't a sign that AI had stopped working. The tool and the task were mismatched.
Some days you want the assistant who knows you broadly and connects ideas across your life. More often than you'd think, you want a clean desk and one folder.
Knowing which one you need is most of what being capable with AI actually means.
ChatGPT isn't AI. It's one doorway into AI.
NotebookLM is another — and it's the one you want when you need the technology to stop wandering.
Consider it decoded.
Want these in your inbox? CLICK HERE — AI decoded for midlife women, straight to you.