Why you forget what you learn

Think about school for a moment.

You probably remember certain things from 10, 20 or even 30 years ago.

Maybe you remember that 3 × 3 = 9.

Maybe you remember the capital of your state.

Maybe you remember the scientific name of something you studied repeatedly.

Maybe you remember a song you haven’t heard in 15 years.

Then there are hundreds of other things you studied that have completely disappeared.

Why?

Repetition matters.

Retrieval matters.

Use matters.

Context matters.

And your brain has limits.

Hermann Ebbinghaus’ early work on forgetting became famous for showing how memory falls away when learned material isn’t revisited. Later research has successfully replicated key features of his classic forgetting curve.

The important part isn’t memorising a particular percentage.

The useful lesson is simpler:

If you don’t revisit information, your ability to retrieve it tends to weaken.

That explains something we have all experienced.

You study something for 2 hours.

At the end, you feel like you know it.

The next morning, you can still explain it.

A week later, somebody asks you a question.

You know you’ve seen it.

You can almost see the page.

You remember the colour of the diagram.

You remember where the paragraph was.

But the actual answer?

Gone.

That feeling is brutal.

And it gets worse when you’re learning as a working adult.

Because you’re not studying one subject for an exam.

You’re learning across your entire life.

Leadership.

AI.

Finance.

Communication.

Technology.

Health.

Business.

Psychology.

Your industry.

Your clients.

Your team.

Your family.

Your own interests.

There is no final exam.

There is just Tuesday morning when you suddenly need the thing you learned 4 months ago.

That’s a different learning problem.

Information is everywhere. Retrieval is the problem.

Open your phone.

How many saved videos do you have?

How many unread newsletters?

How many bookmarks?

How many PDFs?

How many books?

How many screenshots?

How many notes?

How many courses?

How many browser tabs?

You probably have enough information to become reasonably competent in 10 different subjects.

And yet you may still feel like you’re behind.

That’s because collecting information feels like progress.

Sometimes it is progress.

Often, it is only exposure.

You watched the video.

You read the chapter.

You listened to the podcast.

Your brain recognised the information.

Recognition feels good.

Retrieval is harder.

Try this.

Close the book.

Close the video.

Put your notes away.

Now explain the idea from memory.

That’s a different task.

Research on retrieval practice has repeatedly found that actively retrieving information can improve long-term retention compared with simply studying the same material again. Roediger and Karpicke’s work is one of the well-known foundations for this finding, and later research continues to examine how retrieval practice supports retention and transfer.

So when you want to remember what you learn, reading more isn’t always the answer.

Sometimes you need to stop reading.

And try to remember.

That little moment of struggle is doing useful work.

The strange productivity trap of learning

Here’s something I’ve noticed.

Learning can become procrastination.

You watch another video.

You save another article.

You buy another book.

You join another course.

You download another PDF.

You tell yourself:

“I’m learning.”

And technically, you are.

Yet nothing changes.

You haven’t used the idea.

You haven’t explained it.

You haven’t tested it.

You haven’t connected it to another idea.

You haven’t made a decision with it.

You haven’t built anything with it.

The information stays outside you.

That’s the gap between information and knowledge.

Knowledge becomes useful when you can retrieve it and do something with it.

And that is where I think AI can change the way we learn.

Cognitive load is quietly eating your attention

John Sweller’s Cognitive Load Theory gives us another useful piece of the puzzle.

Working memory has limits.

When too much information competes for your attention at the same time, learning becomes harder. Cognitive Load Theory has been extensively developed within educational psychology and instructional design, including work by Sweller and later researchers.

Think about what happens when you sit down to learn after a long workday.

You have 19 browser tabs open.

Slack is running.

WhatsApp is buzzing.

Your phone is beside you.

Your email is open.

You’ve got 3 unfinished tasks in your head.

You haven’t slept properly.

Then you open a 90-minute course and tell yourself:

“Right. Time to focus.”

Good luck.

Your brain has already got a queue.

Then you add another stream of information.

And when you can’t concentrate, you blame yourself.

“I’m getting lazy.”

“I’ve lost my discipline.”

“I can’t focus anymore.”

Maybe the problem is simpler.

Your mental desk is already covered in papers.

Now you’re trying to read another book on top of them.

The doomscrolling loop makes learning harder

This part of the transcript matters to me because it’s so familiar.

You finish work.

You’re tired.

You open Instagram or YouTube Shorts.

One video.

Then another.

Then another.

You aren’t really resting.

You’re consuming.

Your brain keeps switching.

Then suddenly 45 minutes have disappeared.

You feel tired.

You haven’t moved forward.

So you watch another video.

It’s a strange loop.

And when you finally sit down to learn something deliberately, sustained attention feels difficult.

That doesn’t mean short-form content is inherently bad.

It means your learning environment matters.

If your attention is constantly being pulled in different directions, asking yourself to suddenly perform deep learning is a difficult request.

So the first step in a better learning system isn’t an AI tool.

It’s choosing what you are actually trying to learn.

One topic.

One question.

One notebook.

One learning goal.

That’s where we start.

What does it actually mean to remember what you learn?

Let’s make this practical.

Suppose you read a book about public speaking.

You finish it.

What does it mean to say you remember it?

Can you explain the central idea?

Can you recall the 5 most useful concepts?

Can you compare them with another author’s approach?

Can you explain the idea to a 12-year-old?

Can you apply one technique in your next presentation?

Can you answer a question about it without opening the book?

Can you connect it to something you’ve already experienced?

That’s a much better definition of learning.

Remembering is useful when it improves what you can do.

And that changes the workflow.

You don’t simply consume.

You retrieve.

You question.

You connect.

You apply.

You revisit.

AI can become part of your learning system

This is where the idea in the original video gets interesting.

A lot of people use AI like this:

Question.

Answer.

Next question.

Next answer.

Close the chat.

The information disappears into the conversation.

You may have learned something.

You may also forget it tomorrow.

The more interesting approach is to give AI your actual learning materials.

Your book.

Your notes.

Your meeting slides.

Your PDFs.

Your YouTube videos.

Your previous writing.

Your research.

Your own reflections.

Now the AI has something specific to work with.

NotebookLM is built around this source-grounded approach. Google describes it as an AI research assistant where you can upload sources such as PDFs, websites, YouTube videos, audio files, Google Docs and Slides, then chat with the notebook using those sources.

And this distinction matters.

You aren’t simply asking:

“Tell me about public speaking.”

You’re asking:

“Based on these 3 books, these 2 courses and my notes, explain the differences between the authors’ approaches to handling nervousness before a presentation.”

That’s a completely different learning experience.

Your first notebook should have ONE topic

This is one of the simplest ideas in the transcript.

Choose one topic.

Only one.

It could be:

• Public speaking
• Python
• Investing
• Nutrition
• Project management
• Leadership
• Negotiation
• AI
• Writing
• Communication

Don’t start by creating a notebook called:

“Everything I want to learn.”

That becomes a digital junk drawer.

Give the notebook a job.

For example:

Public speaking

Then collect the material that belongs to that learning goal.

That’s how you make it useful.

Google’s current documentation describes each notebook as an independent collection of sources for a specific project. A notebook can’t access information across your other notebooks at the same time.

That actually fits the learning model nicely.

One notebook.

One subject.

One context.

One learning journey.

Start with 3 resources

The transcript suggests beginning with only 3 resources.

I like that.

You don’t need 47 documents on day one.

Start small.

For a public speaking notebook, you might upload:

1. One book
2. One strong YouTube course
3. Your own notes

That’s enough to start asking questions.

You can add more later.

NotebookLM currently supports a much wider range of sources than the original workflow described, including PDFs, Google Docs, Slides, Sheets, Word documents, text, Markdown, images, audio, web URLs, ePub files and public YouTube URLs.

That means your learning library can grow with you.

But don’t confuse capacity with a reason to upload everything.

A clean library is easier to learn from than a messy one.

Build a notebook about yourself

This is one of my favourite examples from the transcript.

The notebook called About Surya wasn’t built around a conventional academic subject.

It was built around a person.

The sources included:

• A LinkedIn profile PDF
• Information about work experience
• Certifications
• Education
• A master prompt containing answers about thoughts, perspectives and habits
• Multiple interviews
• A voice profile interview

Think about what that creates.

You have a collection of information about yourself.

Now you can ask questions against that collection.

You can ask:

“Explain my professional background like I’m 12.”

“What patterns do you see in my career?”

“What are the strongest themes across my experience?”

“What contradictions appear in my answers?”

“What kind of communicator do I appear to be?”

“Based on my own material, what are my strongest areas?”

The answers become more personal because the system has personal source material to work with.

That’s the important idea.

Personalisation starts with personal information.

Why source-grounded AI feels different

Let’s use a simple example.

You ask a general AI:

Explain Surya like I’m 12 years old.

The model has to construct an answer from whatever information it has available in that conversation and its broader model knowledge.

Now imagine you give a notebook:

• Your LinkedIn profile
• Your CV
• Your interviews
• Your own answers
• Your voice profile
• Your writing samples

Then ask the same question.

The second system has a much richer source base.

NotebookLM’s current chat documentation says its standard chat responses use information from the sources in the notebook, with inline citations that let you inspect where an answer came from.

That doesn’t mean the output is automatically correct.

AI can still make mistakes.

Google itself tells users to check NotebookLM outputs.

But the source grounding gives you something extremely useful:

A trail back to the material.

That’s important when you’re learning.

You can ask:

“Where did you get that?”

And inspect the source.

Don’t treat AI as the memory itself

This is where I would make one adjustment to the original idea.

AI can become a very useful learning assistant.

It doesn’t replace your memory.

It doesn’t do the learning for you.

And NotebookLM itself doesn’t make practice unnecessary.

The transcript says this clearly.

You still have to implement.

Your knowledge compounds through use.

That’s why I wouldn’t build a system designed to let AI remember everything instead of you.

I’d build one that helps you remember what you learn.

That’s a subtle difference.

The AI stores the library.

You build the understanding.

The AI asks questions.

You retrieve the answer.

The AI compares sources.

You decide what matters.

The AI organises information.

You apply it.

That’s a much healthier relationship with the tool.

Your 12-year-old test

The transcript uses one beautifully simple prompt:

“Explain this topic to me like I’m 12 years old.”

Use it.

It forces the system to explain things clearly.

It also gives you a useful learning test.

If you can’t understand the explanation, ask another question.

If you understand it but can’t explain it yourself, retrieve it.

If you can explain it but can’t use it, practise.

If you can use it, you’ve crossed another important line.

You can also turn the prompt around.

Ask:

“Explain this concept simply, then ask me 5 questions to test whether I actually understand it.”

Now you have retrieval practice.

And that’s much closer to real learning.

The difference between passive and active learning

The transcript draws a strong distinction between passive and active learning.

Passive learning is when somebody else controls the pace.

They teach.

You listen.

You take notes.

You memorise.

You repeat.

Active learning starts with your curiosity.

You ask questions.

You challenge ideas.

You test assumptions.

You compare sources.

You try to explain something.

You apply it.

You make mistakes.

You ask another question.

That’s a completely different experience.

And AI is unusually good at one part of this.

It doesn’t get tired of your questions.

You can ask the same thing 5 different ways.

You can ask a stupid question.

You can ask a very advanced question.

You can ask:

“Why?”

Then:

“Why?”

Then:

“Why?”

Again.

And again.

No teacher is standing there looking at the clock.

Build your own question engine

Once your notebook has the sources, don’t stop at:

“Summarise this.”

That’s probably the least interesting thing you can do.

Ask questions that force thinking.

Try:

What are the 5 most important ideas across these sources?

Then:

Which ideas disagree with each other?

Then:

Explain why the authors disagree.

Then:

Give me a real-world example of each position.

Then:

Ask me 10 questions to test whether I understand the differences.

Then:

Identify where my answers are weak.

Then:

Give me one practical exercise to improve the weakest area.

That’s a learning loop.

And it starts to look like a personal tutor.

Compare authors instead of collecting authors

Let’s say you’ve read 3 books on leadership.

You could ask NotebookLM to summarise each one.

Fine.

A better question might be:

Compare the 3 authors’ definitions of effective leadership. Where do they agree? Where do they disagree? What evidence does each use?

Now you’re creating connections.

That’s where learning gets interesting.

You’re no longer storing 3 separate books.

You’re building a mental map between them.

The notebook becomes a place where ideas meet.

Ask the AI to challenge you

Here’s another prompt worth saving:

Based only on the sources in this notebook, ask me 10 progressively harder questions about this topic.

Do not show me the answers immediately.

Wait for my response to each question.

After I answer, tell me:
1. What I got right
2. What I missed
3. What I misunderstood
4. Which source supports the correction
5. One follow-up question I should answer

Now you’re doing retrieval.

You’re getting feedback.

You’re returning to the sources.

And you’re making your brain work.

That’s a far better use of AI than asking for another summary.

Retrieval beats another summary

This deserves its own section because it changes everything about how you use learning tools.

Imagine you’ve read a chapter.

You have 2 choices.

Option 1

Ask AI:

“Summarise the chapter.”

You read the summary.

You feel productive.

Option 2

Ask AI:

“Ask me 10 questions about the chapter. Don’t show me the answers until I respond.”

You struggle.

You get some wrong.

You realise you misunderstood one concept.

You go back.

You retrieve it again.

Which one feels easier?

The first.

Which one is more demanding?

The second.

And that difficulty is useful.

Research on the testing effect has repeatedly found that retrieval can improve later retention more than repeated studying, particularly after a delay.

So if your goal is to remember what you learn, build retrieval into the workflow.

Revision should become a conversation

Revision doesn’t have to mean reading the same chapter again.

You can turn revision into a conversation.

Imagine you’ve learned about negotiation.

On Monday:

You read the material.

On Tuesday:

NotebookLM asks you questions.

On Friday:

You explain the concepts in your own words.

Next week:

You apply one concept to a real conversation.

Two weeks later:

You ask the notebook:

“What are the 5 ideas I haven’t used yet?”

Now revision has become active.

You are moving between:

Exposure → Retrieval → Connection → Application → Review

That is a much richer learning cycle.

Research on spaced learning and retrieval supports revisiting material across time rather than relying entirely on one long study session.

The 30-minute learning system

The transcript opens with a powerful promise:

You can build a learning workflow that takes around 30 minutes a day.

The point isn’t that 30 minutes is some universal scientific requirement.

It isn’t.

Your schedule and subject will determine what makes sense.

The useful idea is to create a small daily learning habit that includes retrieval and action.

Here’s the workflow I’d build from the transcript.

Minutes 1 to 5: recall

Before opening your sources, ask yourself:

“What do I remember from yesterday?”

Write it down.

No cheating.

No notebook.

No AI.

Just your memory.

Minutes 5 to 10: check

Open your notebook.

Ask:

“Compare my answer with the sources. What did I get right, what did I miss and what did I misunderstand?”

Now you have feedback.

Minutes 10 to 20: learn

Choose one question.

Only one.

Ask NotebookLM to explain the answer using your sources.

Then ask:

“Give me a real example.”

Then:

“Give me an example from my work.”

Now the concept has context.

Minutes 20 to 25: connect

Ask:

“How does this idea connect to the other concepts in this notebook?”

Then:

“Which previous idea does this challenge or extend?”

Now you’re building relationships between ideas.

Minutes 25 to 30: apply

Finish with:

“Give me one practical action I can take today using this concept.”

Then do it.

That’s the important part.

Do the thing.

Your learning loop should end in action

This is one of the strongest points in your transcript.

NotebookLM doesn’t make practice disappear.

You still have to implement.

If you learn a negotiation technique, use it.

If you learn a communication technique, practise it.

If you learn a programming concept, write code.

If you learn a management idea, try it with your team.

If you learn a public speaking technique, use it in your next presentation.

That’s how information moves towards knowledge.

The final question should always be:

“Where can I use this?”

Use NotebookLM for meeting preparation

This is one of the most practical applications.

Imagine you have a meeting tomorrow.

You have:

• 20 slides
• 3 previous meeting notes
• A project document
• Several emails
• A proposal
• A spreadsheet

Put the relevant sources into a notebook.

Then ask:

“Give me the 5 decisions we need to make in this meeting.”

Then:

“What unresolved questions appear across these sources?”

Then:

“What are the strongest arguments for each option?”

Then:

“What information is missing?”

Then:

“Give me a 5-minute briefing before I enter the meeting.”

Now you’re learning from the material while preparing for something real.

That’s where the system becomes useful.

Use it for a book

Let’s say you want to read a book about leadership.

Upload the book.

Then ask:

“What is the central argument?”

Then:

“What are the 7 ideas the author returns to repeatedly?”

Then:

“Explain each idea with an example.”

Then:

“Which ideas are practical for a manager?”

Then:

“Ask me 10 questions to test whether I understood the book.”

Then:

“Create a 7-day implementation plan based only on this book.”

Now you aren’t simply consuming the book.

You’re interacting with it.

And you have a path from reading to implementation.

Use it for YouTube learning

The current NotebookLM documentation confirms that public YouTube videos can be added as sources.

That creates an interesting workflow.

You find a 90-minute course.

Add it to your notebook.

Then ask:

“What are the main concepts taught in this video?”

Then:

“Which ideas have practical demonstrations?”

Then:

“Create 10 retrieval questions from the video.”

Then:

“Which concepts should I practise first?”

Then:

“Explain the most difficult concept in simple language.”

You have turned a video into an interactive learning resource.

Use it for PDFs you already have

Think about the PDFs sitting on your laptop.

Research papers.

Company reports.

Course notes.

Industry reports.

White papers.

Conference material.

Product documentation.

Instead of reading every document from beginning to end immediately, you can use your notebook to find the parts relevant to your current question.

NotebookLM can work with PDF files and many other document formats.

Try:

“Find the sections related to customer retention.”

Then:

“Compare the recommendations across these 4 documents.”

Then:

“Which claims are supported by multiple sources?”

Then:

“Where do the sources disagree?”

That turns your document collection into something you can interrogate.

Use research to grow the notebook

The transcript also demonstrates using research to expand a notebook.

For example:

“Find the greatest TEDx talks and full YouTube courses on public speaking and communication skills.”

The current version of NotebookLM has Fast Research and Deep Research capabilities for discovering and importing sources. Google says Deep Research can browse large numbers of websites, create a report and let you import the report and sources into a notebook.

That means your notebook can grow.

You start with 3 sources.

You learn.

You discover gaps.

You research those gaps.

You add better sources.

You ask new questions.

You update your understanding.

That’s a living learning system.

Be careful with the word “research”

There’s one thing I’d change from the original transcript.

Don’t treat AI research as automatically authoritative.

If NotebookLM finds 20 websites, that doesn’t make all 20 correct.

You still need source judgement.

Look at:

Who wrote it?

When was it published?

What evidence does it use?

Is the source primary or secondary?

Does another credible source agree?

Does the claim actually appear in the cited source?

NotebookLM’s citations make checking easier, which is one of its strengths. Google specifically describes inline citations that allow you to inspect the supporting source material.

Use that feature.

Don’t outsource judgement.

Your notebook can become a personal knowledge hub

Now zoom out.

Imagine doing this for every major area of your life.

One notebook for leadership.

One for AI.

One for public speaking.

One for investing.

One for your business.

One for writing.

One for a major project.

One for a professional certification.

Each notebook has its own sources.

Each notebook becomes a focused learning environment.

Google’s current system treats notebooks as separate collections, which fits this structure well.

And over time, your source library grows.

Your notes grow.

Your questions grow.

Your understanding grows.

You now have somewhere to return to.

That matters.

Because one of the biggest problems with learning is that we often throw away the context after we’ve consumed the information.

A notebook keeps the context.

Add your own notes

There is another important feature here.

Your own thinking should become part of the system.

NotebookLM lets you create notes, save useful chat responses as notes and convert notes into sources. Google currently allows up to 1,000 notes per notebook.

So after a learning session, write:

What I learned

What surprised me

What I disagree with

What I want to test

Where I can apply this

Those notes are yours.

They show how your understanding is changing.

And that creates something much more interesting than a document archive.

It creates a record of your thinking.

Your learning system should know what you think

This is why the About Surya example is so powerful.

The notebook contains information about Surya.

But it also contains his answers to questions about:

• Thought process
• Perspectives
• Habits
• Interviews
• Voice
• Experience

That means the notebook can respond within a context that belongs to him.

You can build the same idea around your own learning.

Suppose you’re studying leadership.

Add:

• Your leadership notes
• Feedback you’ve received
• Examples from your career
• Books
• Course notes
• Meeting reflections
• Difficult situations
• Your own principles

Then ask:

“What leadership patterns keep appearing in my experiences?”

That’s where learning becomes personal.

Turn your mistakes into sources

This is another habit worth building.

After something goes wrong, document it.

A presentation didn’t land.

Write what happened.

A project slipped.

Write why.

A client objected.

Write the conversation.

You made a decision that turned out badly.

Record the decision and the result.

Then put those reflections into your learning system.

Later, ask:

“What patterns appear across my mistakes?”

That’s powerful.

Because your own life becomes part of the curriculum.

Steve Jobs and the dots

The transcript connects this idea with Steve Jobs’ famous Stanford commencement speech.

Jobs talked about connecting the dots backwards, looking at experiences later and seeing how they connect.

That idea fits beautifully here.

Because a learning system gives you somewhere to store those dots.

You read something.

You try something.

You fail.

You learn.

You meet someone.

You discover an idea.

You change your approach.

Months later, those things start connecting.

AI can help you inspect those connections.

It can’t manufacture meaning for you.

You still decide what the experience means.

But it can help you find patterns you might have missed.

Ask your notebook to find the dots

Try this:

Review the sources in this notebook and identify recurring ideas, experiences, beliefs or patterns.

Group related ideas together.

For each pattern:
1. Show the sources that support it
2. Explain why they appear related
3. Identify any contradictions
4. Give me one question I should explore further

Do not invent connections that aren't supported by the sources.

That’s a much more interesting use of AI.

You’re asking it to help you see relationships across your own material.

What happens when your learning library gets bigger?

This is where the system becomes increasingly useful.

At first, you have 3 sources.

Then 10.

Then 20.

Then 50.

You have years of notes.

Your own reflections.

Books.

Videos.

Courses.

Reports.

Projects.

Now the question changes.

Instead of:

“Where did I save that?”

You can ask:

“Which of my previous notes relate to this problem?”

Instead of:

“What did that book say?”

You can ask:

“How does that author’s argument compare with what I learned last year?”

Instead of:

“What do I know about this?”

You can ask:

“What gaps exist in my current understanding?”

That is where your learning library becomes genuinely useful.

The danger of building a giant digital junk drawer

There is a catch.

More sources aren’t automatically better.

Imagine putting:

500 PDFs

1,000 videos

3,000 notes

200 books

and every random article you’ve ever saved

into one notebook.

Now you’ve built a monster.

Your notebook needs structure.

Use separate notebooks for separate learning areas.

Use clear names.

Use notes.

Use source selection.

NotebookLM lets you select or deselect sources when chatting, which gives you more control over what the model uses for a particular question.

That matters when your notebook grows.

Sometimes you want the entire library.

Sometimes you want 3 sources.

Choose deliberately.

Your ideal source stack

For a new topic, I’d start with:

Source 1: primary material

A book, course, paper or official document.

Source 2: explanation

A strong lecture, interview or YouTube course.

Source 3: your own notes

Your questions.

Your understanding.

Your context.

Then expand.

Add another expert.

Add a contrasting perspective.

Add a practical case study.

Add your implementation notes.

Now you have a richer learning environment.

The question ladder

One of the easiest ways to get more from your notebook is to ask questions in levels.

Level 1: understand

“What does this mean?”

Level 2: simplify

“Explain this like I’m 12.”

Level 3: compare

“How does this differ from X?”

Level 4: challenge

“What assumptions does this argument make?”

Level 5: retrieve

“Ask me questions without showing the answers.”

Level 6: apply

“Where could I use this tomorrow?”

Level 7: reflect

“What does this change about how I currently work?”

Level 8: create

“Help me build something using this idea.”

That’s a learning ladder.

You don’t have to climb every rung every day.

But it gives you somewhere to go after the first summary.

The 5 questions I would ask every day

If you want this to stay simple, use these.

1. What did I learn?

Retrieve it from memory.

2. What did I get wrong?

Check against the source.

3. What does it connect to?

Build relationships.

4. Where can I use it?

Find an application.

5. What should I remember next week?

Create the next retrieval point.

That’s enough to change your learning habit.

What AI should do in your learning system

Give AI the boring work.

Let it:

• Search your sources
• Find relevant passages
• Compare authors
• Explain difficult ideas
• Generate questions
• Create study guides
• Organise notes
• Identify patterns
• Help you prepare
• Create revision material

NotebookLM currently supports study guides, briefings, Audio Overviews, mind maps, flashcards, quizzes, infographics and slide decks, among other outputs.

That’s a lot.

But your job remains:

Think. Retrieve. Decide. Apply.

That’s where the learning happens.

Don’t let AI make learning too easy

This sounds strange.

You have a tool that can explain almost anything.

Why would you deliberately make learning harder?

Because effort matters.

If AI answers every question before you have tried to answer it, you’re training yourself to ask rather than remember.

Try this instead.

First:

“What do I think?”

Then ask AI.

First:

“Can I explain this?”

Then check.

First:

“What are the 3 reasons?”

Then compare with the source.

Give your brain a chance.

The blank-page retrieval test

Here’s a simple exercise.

After learning a topic, open a blank document.

Set a timer for 5 minutes.

Write everything you remember.

No notes.

No AI.

No searching.

Just write.

Then open your notebook.

Compare.

You’ll probably discover something interesting.

You remembered the broad idea.

You forgot the details.

You remembered one example.

You forgot the mechanism.

You remembered the conclusion.

You forgot the evidence.

That’s useful.

Now you know what needs another pass.

The teach-back test

Now explain the concept out loud.

Pretend you’re explaining it to a 12-year-old.

No slides.

No notes.

No fancy terminology.

If you get stuck, you’ve found a gap.

Go back to the source.

Ask NotebookLM to explain that specific part.

Then try again.

This is active learning.

You’re doing something with the knowledge.

The implementation test

There is an even harder test.

Use it.

Suppose you’ve learned:

“Good meetings need clear decisions.”

Fine.

Your next meeting arrives.

Do you actually ask:

“What decision are we making today?”

That’s where learning becomes behaviour.

And behaviour is where the value appears.

You can have perfect notes about leadership and still manage badly.

You can know 20 productivity methods and still procrastinate.

You can understand communication theory and still avoid the difficult conversation.

The gap is implementation.

Your learning workflow should deliberately cross it.

The compounding effect

Let’s say you learn one useful concept every day.

That’s 365 concepts in a year.

But the real benefit isn’t 365 isolated facts.

It’s what happens when those ideas connect.

Leadership connects with psychology.

Psychology connects with communication.

Communication connects with negotiation.

Negotiation connects with business.

Business connects with technology.

Technology connects with learning.

Learning connects back to leadership.

Your knowledge starts forming a network.

That’s why meaningful connections matter.

Research on retrieval also suggests that retrieval can support later flexible access and transfer of knowledge across contexts.

Learning compounds when ideas become useful in more than one situation.

Who should build this kind of learning system?

The transcript names several groups.

Lifelong learners.

Professional managers.

Content creators.

Consultants.

Founders.

Business owners.

Busy professionals who learn from many sources but don’t have those sources organised.

I’d add one simple filter.

If your work requires you to keep learning, this is worth exploring.

You don’t need to be a student.

In fact, working professionals may have an even stronger use case.

You aren’t learning for a test.

You’re learning because tomorrow’s work requires something you don’t know today.

Your personal learning library

Imagine your notebook system 12 months from now.

You have:

Books

A record of the ideas you’ve studied.

Videos

Courses and talks you’ve learned from.

Notes

Your own interpretation.

Questions

The things you’re still trying to understand.

Projects

Where you applied the knowledge.

Reflections

What happened.

Mistakes

What didn’t work.

Research

What you discovered later.

Now your learning system has history.

You can return to it.

Ask questions.

Find patterns.

Refresh your memory.

Build from previous work.

That’s much more useful than another pile of bookmarks.

A simple notebook structure

You don’t need anything complicated.

Use this:

Sources

Your books, videos, PDFs, documents and notes.

Questions

What you’re trying to understand.

Learnings

Important ideas in your own words.

Applications

Where you used them.

Reflections

What happened after you tried them.

Review

What you need to revisit.

That’s enough.

The weekly review

Once a week, spend 30 minutes reviewing your learning.

Ask your notebook:

“What did I learn this week?”

Then:

“Which ideas did I actually apply?”

Then:

“Which ideas did I forget?”

Then:

“Which ideas connect with things I’ve learned previously?”

Then:

“What should I revisit next week?”

This gives your learning system a rhythm.

You aren’t relying on memory to tell you what you forgot.

You’re deliberately checking.

The monthly review

Once a month, zoom out.

Ask:

“What were the most useful ideas I learned this month?”

Then:

“Which ideas changed my behaviour?”

Then:

“Which ideas did I learn but fail to apply?”

Then:

“What patterns appear across my learning?”

Then:

“What should I stop learning because it isn’t relevant to my current goals?”

That last question is underrated.

Your attention is limited.

Choose where it goes.

Build a learning system around a current goal

This is another point worth making.

Don’t learn everything.

Learn what you currently need.

If you’re preparing for a promotion, build around leadership.

If you’re launching a business, build around the knowledge required for that business.

If you’re becoming a better speaker, build around communication.

If you’re learning Python for a project, build around Python.

The topic gives the system direction.

Your goal gives the topic meaning.

What about the “second brain”?

The transcript uses the phrase second brain.

I like the metaphor when we keep it grounded.

Your second brain shouldn’t mean:

“AI thinks for me.”

It should mean:

“I have an external system that stores, organises and helps me work with knowledge.”

The brain remains yours.

Your judgement remains yours.

Your decisions remain yours.

The external system gives you somewhere to put information so your biological brain doesn’t have to hold every detail at once.

That can be incredibly useful.

The second brain should reduce cognitive clutter

Remember the overloaded desk?

Your brain doesn’t need to remember where every PDF lives.

It doesn’t need to remember which book contained the quote.

It doesn’t need to remember which meeting note had the decision.

It doesn’t need to remember every detail from every course.

Put those things somewhere you can retrieve them.

Then spend your mental energy on:

Understanding.

Connecting.

Deciding.

Creating.

Applying.

That’s the point of an external knowledge system.

NotebookLM is a tool. Your workflow is the system.

This is probably the biggest takeaway from the entire transcript.

You can open NotebookLM tomorrow.

Upload 3 PDFs.

Ask 5 questions.

Close it.

Nothing changes.

The tool didn’t fail.

The workflow never existed.

The real system looks like:

Choose → Collect → Question → Retrieve → Connect → Apply → Review

That’s the loop.

AI sits inside it.

Your first notebook, step by step

Let’s make this extremely practical.

Step 1: Pick one topic

Choose something you genuinely want to learn.

For example:

Public speaking

Step 2: Create the notebook

Google’s current NotebookLM documentation says you can create a new notebook and add sources directly when you begin.

Give it a clear name.

For example:

Public speaking learning system

Step 3: Add 3 sources

Choose:

1. One book
2. One strong video or course
3. Your own notes

Keep it small.

Step 4: Ask for a simple explanation

Use:

Explain this topic to me like I'm 12 years old.

Use only the sources in this notebook.

Give me:
1. The central idea
2. The 5 most important concepts
3. One real-world example for each
4. The biggest mistake a beginner makes
5. One thing I should practise today

Step 5: Retrieve

Close the notebook.

Write what you remember.

Then check.

Step 6: Ask questions

Use:

Ask me 10 questions that test whether I actually understand this topic.

Start easy and become progressively harder.

Do not show the answers until I respond.

Step 7: Find connections

Ask:

How do the major ideas in these sources connect?

Identify:
1. Agreements
2. Disagreements
3. Dependencies
4. Contradictions
5. Practical connections

Step 8: Apply

Ask:

Based only on these sources, give me one practical exercise I can complete today.

Make it specific enough that I can do it immediately.

Then do it.

Step 9: Add your reflection

Write:

What happened?

What worked?

What didn’t?

What surprised me?

What will I change?

Add the note to your notebook.

Step 10: Review later

Come back tomorrow.

Retrieve.

Then again next week.

That’s your learning loop.

A complete prompt pack for your notebook

You can save these prompts.

Understand

Explain this topic like I'm 12 years old.

Use only the sources in this notebook.

Give me the simplest accurate explanation first, then add the important details.

Compare

Compare the main ideas across these sources.

Show where they agree, where they disagree and why the difference matters.

Cite the relevant sources.

Retrieve

Test me on this topic.

Ask one question at a time.

Do not reveal the answer before I respond.

After each answer, tell me what I got right and what I missed.

Apply

Give me one practical situation where I can apply this concept today.

Then give me a specific action to take.

Connect

Find connections between this topic and the other ideas in this notebook.

Only include connections supported by the sources.

Show the evidence for each connection.

Review

Help me review this topic.

Ask me what I remember first.

Then compare my answer with the sources and identify the gaps.

Challenge

Challenge my understanding.

Give me 5 questions that require application rather than memorisation.

Wait for my answer before giving feedback.

What changes when you stop trying to remember everything?

This is where the idea becomes bigger than NotebookLM.

You don’t need to keep every detail inside your head.

You need to know:

What matters?

Where is it?

How do I retrieve it?

How do I use it?

That’s a much more realistic goal.

Your external system remembers the location and context.

Your brain develops understanding.

Your practice develops skill.

Your experience develops judgement.

Together, those things become useful knowledge.

The real enemy is disconnected information

Think about 10 books sitting separately on a shelf.

They contain knowledge.

But they aren’t talking to each other.

Now imagine you could ask:

“How does the idea in Book 3 challenge the idea in Book 7?”

Or:

“What did I learn from these books that applies to the problem I’m facing today?”

Now the library becomes interactive.

That’s the promise of a source-grounded AI notebook.

Not that the machine magically knows everything.

That it can help you work across the material you’ve given it.

Remember what you learn by turning learning into a loop

This is the whole thing.

You read.

Then retrieve.

You ask.

Then explain.

You compare.

Then connect.

You apply.

Then reflect.

You revisit.

Then repeat.

That’s how you stop treating learning as consumption.

And that’s how you start building a personal learning practice.

The 30-minute daily template

Save this somewhere.

0 to 5 minutes

Recall

Write everything you remember from yesterday.

5 to 10 minutes

Check

Use your notebook to find gaps.

10 to 15 minutes

Question

Ask one difficult question.

15 to 20 minutes

Connect

Relate the idea to something you already know.

20 to 25 minutes

Retrieve

Answer 3 to 5 questions without looking.

25 to 30 minutes

Apply

Do one real-world action.

That’s your daily loop.

You don’t need a complicated productivity system.

You need a repeatable learning habit.

What your AI second brain should eventually contain

If you keep going, your system can contain:

Your knowledge

Books, courses, videos and research.

Your experience

Projects, meetings, decisions and mistakes.

Your thinking

Notes, questions, beliefs and reflections.

Your work

Documents, presentations and reports.

Your learning history

What you’ve studied and what you’ve applied.

Your current goals

What you’re trying to become better at right now.

That gives AI much better context.

And it gives you a place to return when your own memory says:

“I know I learned this somewhere…”

The biggest mistake to avoid

Don’t build the system and forget to use it.

I’ve seen this happen with every productivity tool.

Beautiful workspace.

Perfect folders.

Amazing colour coding.

Then nothing.

The learning system should be slightly boring.

That’s fine.

You open it.

You ask.

You retrieve.

You apply.

You close it.

Tomorrow, you come back.

The value comes from repetition.

Start with the topic you can’t stop thinking about

Don’t begin with the topic you think you should learn.

Begin with the topic you’re genuinely curious about.

Maybe it’s:

AI.

Investing.

Public speaking.

Leadership.

Python.

Nutrition.

Writing.

Your business.

Your next career move.

Pick one.

Then give it a home.

What happens after 90 days?

You might have:

1 notebook.

Then 3.

Then 5.

Your sources become organised.

Your questions become better.

Your notes become richer.

Your retrieval becomes easier.

You start recognising connections.

You begin remembering ideas because you’ve used them.

And something else happens.

You become more selective about what you consume.

Because you realise that information isn’t the goal.

Useful knowledge is the goal.

Frequently asked questions

Does AI actually help you remember what you learn?

AI can support the learning process, especially when you use it for explanation, questioning, retrieval practice, source comparison and review.

The evidence for retrieval practice itself is strong. Research has found that retrieving information can improve later retention compared with repeated study alone.

AI doesn’t remove the need to retrieve.

It can make the retrieval process easier to organise.

Is NotebookLM the same as ChatGPT?

They can overlap in what they can do, but they are designed around different workflows.

NotebookLM is strongly centred on your selected sources. Google says standard NotebookLM chat responses are grounded in the sources within the notebook and provide citations to those sources.

ChatGPT can also work with supplied files and context, depending on the product and setup.

The useful distinction here is the workflow.

A dedicated notebook gives you a persistent collection of sources around a particular learning goal.

Can NotebookLM prevent hallucinations?

Don’t treat it as a guarantee.

Grounding responses in your sources can reduce unsupported answers and makes checking easier, but AI can still make mistakes.

Google explicitly advises users to double-check NotebookLM outputs.

For anything important, check the cited source yourself.

How many sources should I upload?

The transcript recommends starting with 3.

That’s a good starting point.

You can add more as your questions become more specific.

The current product supports many source types and allows substantially more sources depending on your account and plan.

The number matters less than relevance.

Can I use YouTube videos?

Yes.

Google currently lists public YouTube URLs among the supported NotebookLM source types.

That makes it particularly useful for turning video-based learning into something you can question and review later.

Can I use PDFs?

Yes.

PDFs are among the supported source types.

You can use books you have the right to upload, research papers, reports, course material and other relevant documents.

Respect copyright when adding material.

Should I use AI to summarise every book?

I’d resist that habit.

A summary can help you understand the structure of a book.

Then ask questions.

Retrieve.

Compare.

Apply.

The goal is to create interaction with the material.

A pile of summaries can become another pile of information.

How long should I study every day?

There isn’t a universal number.

The 30-minute structure here is a practical starting point from the workflow in the transcript.

Your actual schedule can be shorter or longer.

The important part is the cycle:

Recall → Learn → Connect → Retrieve → Apply → Review

What if I forget anyway?

You will.

That’s normal.

The goal isn’t perfect biological memory.

The goal is to make forgetting predictable and manageable.

When you revisit material, retrieve it and apply it, you give yourself more opportunities to strengthen access to that knowledge.

And when you keep your sources organised, you also make it easier to return when you need them.

Your learning system starts with one notebook

You don’t need to redesign your entire life tonight.

Pick one topic.

Create one notebook.

Add 3 useful sources.

Ask:

“Explain this to me like I’m 12 years old.”

Then close the notebook.

Write what you remember.

Open it again.

Check yourself.

Ask questions.

Compare the sources.

Find a connection.

Use one idea.

Come back tomorrow.

That’s it.

Then do it again.

And again.

Eventually, your notebook becomes more than a place where documents live.

It becomes a place where your questions live.

Your notes.

Your reflections.

Your mistakes.

Your research.

Your ideas.

Your progress.

And your learning starts becoming something you can return to.

That’s the part I find exciting.

Because the future of learning probably isn’t about consuming more information.

We already have enough.

It’s about building better relationships with the information we choose to keep.

It’s about remembering what matters.

It’s about being able to retrieve an idea when the situation calls for it.

It’s about taking something you read on Monday and using it on Thursday.

It’s about looking back at an experience 2 years later and finally seeing the connection.

AI can help with that.

But the human still has to do the most interesting part.

You have to think.

You have to question.

You have to remember.

You have to practise.

You have to use the idea.

And then you have to come back and ask a better question.

That’s how information becomes knowledge.

That’s how knowledge becomes skill.

And that’s how learning starts changing the way you live and work.

Start here today

Choose ONE topic.

Create ONE notebook.

Add 3 sources.

Then ask these 5 questions:

1. What are the most important ideas here?
2. Explain them to me like I’m 12.
3. Ask me 10 questions without showing the answers.
4. What connections exist between these ideas?
5. What can I apply today?

Then close the screen.

Go do the thing.

Tomorrow, try to remember what you learned today.

That’s the real test.

Because the goal was never to consume another book, another course or another AI conversation.

The goal is to remember what you learn when your life actually asks you to use it.

And now you have a system for doing exactly that.

If you want to build the deeper version of this system, start with the complete guide to building your AI second brain.

Then go deeper into the forgetting curve and how retrieval improves learning.

And when you’re ready to turn what you’ve learned into a repeatable daily practice, use this practical learning workflow.

Your next book is waiting.

Your next idea is waiting.

Your next question is waiting.

The only thing left is to make sure you still remember it when you need it.

If you liked reading this Blog, check out the step by step hands on implementation course below