Building a Second Brain with AI

Introduction

We process overwhelmingly amount of information everyday

From the moment we wake up, decisions begin immediately. What should we eat? Should we drive or take public transport? Which tasks should be prioritized first? Most of these decisions happen almost automatically.

Technology only increases this load. Phones, laptops, tablets, and online platforms constantly push new information toward us. While this makes life more convenient, it also creates information overload. Useful ideas are easily forgotten because we consume far more than we can properly absorb.

This creates a problem

Many of us enjoy learning new things, yet struggle to retain or apply what we learn consistently. Information is available everywhere, but recalling the right knowledge at the right time becomes difficult.

This is where the concept of a Second Brain becomes valuable.

In this post, I will explain what a second brain is, how it works, how I personally built a second brain with AI tools and the benefits and limitations of using a second brain from my personal perspective.

My Background

I always enjoyed learning new things. Some were theoretical concepts while others were practical skills.

The problem was never learning itself. It was applying and retaining what I learned.

I often consumed useful ideas, understood them, then moved on without using them properly. Over time, many of those ideas faded away. When I finally needed them, I could no longer recall them clearly.

This is natural.

Knowledge that is not applied is easily forgotten. As the saying goes:

What you don’t use, you lose it.

I came across the concept of second brain last year. It sounded practical. In simple terms, it is a method of storing knowledge externally.

However, traditional note-taking also has limitations. When summarizing information manually, it is easy to miss important ideas or overlook useful connections between topics.

Artificial Intelligence helped solve part of that problem.

With AI tools today, information can be summarized, organized, and retrieved much more efficiently. Instead of relying entirely on memory or incomplete notes, useful ideas can now be stored, revisited, and connected more effectively.

In many ways, AI has made learning more accessible than ever before.

The Second Brain

What is a Second Brain

A Second Brain is a system designed to store and organize knowledge so information can be recalled and used more effectively.

The purpose of the Second Brain is to

• recalled ideas at any time
• connected different ideas together
• developed new perspectives
• broaden our insights

Why is Second Brain Important

We consume enormous amounts of information every day. Articles, videos, books, podcasts, and online platforms constantly compete for our attention. While access to information has become easier, retaining and applying that knowledge has become more difficult.

This is where the Second Brain comes in.

Think of it as a personal librarian that stores information based on your preferences that can be retrieved whenever needed.

The purpose of the Second Brain is not to replace human thinking. Its role is to support it.

By storing knowledge externally, the mind becomes more available for deeper thinking, connecting ideas, solving problems, and creating new perspectives.

The CODE Method

Second Brain use CODE framework: Capture, Organize, Distill and Express

Capture

When you are learning about something, it is important to begin with a clear objective. This is the concept of Capture.

It focuses on looking for specific knowledge. This naturally increases engagement.

Learning becomes more effective when guided by questions.

• How do I earn additional income?
• How do I cook a perfect medium rare steak?
• How do I become fluent in AI?

Questions creates direction and looking for answers makes the process more meaningful.

Organize

Once information is captured, it must be Organized.

A common method used in Second Brain system is the PARA framework. PARA divides information into four categories:

1. Project – Information that relates to ongoing work with deadlines.
2. Area – Information that is important and relevant for a long period of time.
3. Resources – Useful references that may become valuable in the future.
4. Archive – Information that is not important but is worth keeping

The purpose is not only storage, it is to make information easier to retrieve and apply when needed.

Distill

After Capture and Organize, the next step is Distill

Distilling means reducing information to its most important ideas. Instead of storing large amounts of raw information, the focus shifts toward extracting the key insights.

This process makes knowledge easier to review, understand and recall later.

A common approach to distilling information includes:

1. Bold all the important sections
2. Highlight the key points
3. Summarize the highlighted points

Express

The final step is Express.

This is the process of sharing what has been learned through writing, discussion, teaching, or creation.

While storing information is useful, true understanding develops when ideas are expressed externally. Sharing shifts the mindset from passive consumer to active creator.

The more you share the more you learned.

Expression can take the form of writing articles, working on projects, creating artifacts or simply discussing ideas

Building a Second Brain with AI Tools

The concept of Second Brain is interesting. However, It can be very time consuming. Distilling information does required a lot of effort.

Artificial Intelligence helps simplify this process.

With AI tools, information can be researched and summarize more effectively.

To build my own AI-assisted Second Brain, I mainly used:

• Perplexity
• NotebookLM
• Gemini

Workflow

The workflow I used for building a Second Brain consists of four stages.

1. Information Research
2. Creating a Library
3. Implementing NotebookLM into Gemini

Information Research

Large Language Models (LLMs) such as Chat GPT and Gemini are trained on large amounts of data.

However, general LLMs still have limitations

Models may generate inaccurate answers and misleading information. This is commonly referred to as AI hallucinations.

This is where Perplexity becomes useful.

Perplexity combines AI chatbot functionality with real-time web research and citations. Instead of only generating answers, it also provides sources that can be verified manually.

This improves reliability during the research stage.

For example, if a user wants to learn more about artificial intelligence, Perplexity can summarize information from multiple sources while providing inline citations so that credibility can be checked. An example of how citations are done is shown below.

Alternatively, sources can be obtained through personal research like watching a YouTube video or reading a book.

Creating a Library

Once resources are collected, they can be stored in the library NotebookLM.

NotebookLM acts as a centralized knowledge library where documents, articles, videos, and notes can be organized into separate notebooks.

A good practice is to keep one notebook focused on one topic. This improves organization while reducing chances of unrelated information affecting responses.

In the example above, NotebookLM is able to generate different types of documents such as reports, slides, infographics as well.

With this, a library of trustworthy resources is created.

Implementing NotebookLM into Gemini

Once enough resources are stored in NotebookLM, they are automatically integrated into Gemini provided that same Google account is used across both platforms.

This allows Gemini to answer questions using the uploaded resources as references rather than relying only on general internet knowledge.

This improves accuracy significantly.

Instead of a general chatbot, Gemini becomes more specialized.

Furthermore, Gems and be created while providing relevant notebooks as resources.

A Gem is a customized AI assistant designed around a specific role, behavior, or purpose. It can also use selected notebooks as its knowledge source.

Creating a Gem for Second Brain

A Gem requires 4 components:

1. Name:
2. Description:
3. Instructions:
4. Knowledge Sources:

The key to create an efficient Gem is to prompt detailed instructions. There are various prompt frameworks that can be used here and there are no right answers to this. It all depends on what the user wants the Gem to do.

Here is an example of the prompt framework that I used:

#Role
What should the model impersonate as? 
This will help tailored the model's structure when it answers your question
#Context
Tell the model about yourself so that the model can answer questions to 
your preferences
#Task
What does the model do?
Example: Gives projects, homeworks, generates learning material
# Constraints
What "rules" should the model follow?
Example: Answer questions based only on the material provided
#Output Format
Example: Chart, Diagram, paragraphs of no more than 300 words, bullet points

Here is an example of a prompt that I’ve used to create a Gem for my Second Brain based on my notebook about AI.

# Role
You are "AI Tutor Prime," an expert AI educator and computer science mentor. Your tone is encouraging, incredibly clear, and patient. You excel at taking highly complex, technical concepts and breaking them down into digestible, beginner-friendly lessons without losing technical accuracy.
# Context
The user is an AI beginner who has curated a specific repository of AI knowledge (transferred from their NotebookLM workspace). They need a dynamic assistant that can seamlessly scale its depth—ranging from answering quick, casual questions to executing deep-dive structural analyses of difficult AI concepts.
# Task
Your objective is to teach and explain AI concepts using *only* the resources provided in your knowledge base.
1. **Adaptive Depth:** If the user asks a simple question, give a direct, crisp answer. If they ask about a complex framework, break it down step-by-step.
2. **Dual-Layer Explanations:** For every conceptual explanation, provide a real-world analogy *and* a simple, cleanly commented Python code snippet (using libraries like NumPy, PyTorch, or basic vanilla Python) to demonstrate the logic in action.
3. **Interactive Learning:** Conclude detailed explanations with a single, highly engaging "Quick Check" question to verify the user's understanding.
# Constraints
- **Source Grounding:** Rely strictly on the uploaded knowledge base documents. If the user asks a question not covered in your documents, say: "That topic isn't covered in your NotebookLM export. Would you like me to use Google Search to look it up, or should we stick to your files?"
- **Code Accessibility:** Ensure all code snippets are beginner-friendly, well-commented, and focus on the core logic rather than complex software engineering syntax.
- **Formatting Guidelines:** Never use dense walls of text. Prioritize scannability using bold text, bullet points, and markdown tables where appropriate. Do not use LaTeX for simple numbers or basic prose; reserve standard markdown for formatting.
# Output Format
For comprehensive concept explanations or deep dives, adhere to this structure:
## ## [Concept Name]
- **The Core Concept:** (A simple, 2-sentence explanation of what it is)
- **Real-World Analogy:** (A relatable, non-tech parallel)
- **How It Works (Deep Dive):** (Step-by-step breakdown of the mechanics based on your files)
- **Code Implementation:** ```python
  # Simple, highly commented code snippet illustrating the concept

Different instructions can completely change the behavior of the Gem. For example, it could act as a teacher, generate exercises, summarize books or even create learning plans.

Once the Gem is connected to the relevant notebooks, it becomes a personalized AI assistant built around the user’s own knowledge system.

Reflections

AI makes the concept of Second Brain far more practical than before. Tasks that once required extensive manual effort such as researching, summarizing, organizing, and recalling information can now be performed much more efficiently.

However, while AI improves the workflow significantly, it also introduces new limitations and responsibilities.

Building an effective Second Brain still requires human judgment, continuous learning, and proper maintenance.

Benefits

Easy to recall storage

One of the biggest advantages of an AI-assisted Second Brain is the ability to store and retrieve information quickly.

I personally enjoy learning through books, YouTube videos, articles, and online courses. The problem is that I often forget useful ideas shortly after consuming them.

With tools such as NotebookLM and Gemini, useful knowledge can now be stored, organized, and recalled much more effectively.

This reduces the mental burden of remembering large amounts of information while making learning more sustainable over time.

Ability to Capture, Distill and Connect information

AI greatly improves the process of capturing and summarizing information.

Traditionally, distilling large amounts of information requires time, concentration, and repeated review. AI helps reduce that workload by summarizing resources much more quickly.

More importantly, it becomes easier to connect ideas together.

Instead of viewing information as isolated pieces of knowledge, AI-assisted systems make it easier to identify relationships between concepts across different topics. This encourages broader thinking and deeper understanding.

More than just a Second Brain

While we use NotebookLM for storing resources, its capabilities extend beyond simple storage. Uploaded resources can be transformed into reports, study materials, infographics or even podcast discussions.

Limitations

Human in the Loop

Even with the help of Perplexity, a user still needs to choose resources carefully.

AI can summarize information efficiently, but it cannot fully replace human judgment. Users still need to evaluate whether sources are reliable, relevant, and accurate.

This means human involvement remains essential throughout the entire process.

For example, someone building a Second Brain around cooking techniques may gather information from videos, articles, and AI-generated summaries. However, without practical experience or foundational knowledge, misleading or incorrect information may still go unnoticed.

Prompting quality also plays a major role.

The usefulness of the system depends not only on the resources collected, but also on how effectively instructions are given to the AI. While AI makes learning easier, it still requires human understanding and critical thinking.

More information on prompting can be read here

Knowledge Cutoff

A Second Brain only knows what has been provided to it.

Unlike general AI models that can search the web in real time, a notebook-based system depends heavily on its existing resources. As a result, outdated information may remain inside the system unless the user updates it regularly.

Maintaining the knowledge base therefore becomes an ongoing responsibility.

This creates an interesting tradeoff. While limiting resources improves accuracy and reduces hallucinations, it also increases the need for continuous maintenance.

English is the main Language

One limitation I personally noticed is that AI systems currently perform more accurately in English than in Thai.

This became especially noticeable when using NotebookLM to summarize Thai audio recordings. The accuracy dropped significantly compared to English resources.

This creates a challenge when building a Second Brain around Thai-language podcasts, videos, or discussions.

A major reason for this is that many AI systems are trained primarily on English-based datasets. Since English dominates much of the internet, models naturally perform better when processing English prompts and resources.

As AI continues improving, this limitation will likely decrease over time. However, language quality still remains an important consideration when building AI-assisted systems today.

Cost Effectiveness

While an AI-assisted Second Brain is powerful, the cost can become difficult to justify.

To maximize the workflow, subscriptions such as Gemini Pro and Perplexity Pro are recommended. Together, they cost roughly $40 per month or $480 per year.

For a simple note storage and summarization, this may feel excessive.

The workflow can still be done with free versions. However, free versions reduces the models capability such as deep research and file storages.

That said, Gemini Pro still provided strong values because of its integration with other Google services such as Gmail, Google Docs, and Google Sheets.

Perplexity Pro, however, depends heavily on the user’s needs. Since Perplexity mainly focuses on research, its value increases significantly only when deep research is required regularly.

Future Improvements

While the current workflow is already useful, there are still several areas I would like to improve.

Most of my current resources come from YouTube videos. In the future, I would like to incorporate more research-based sources through tools such as Perplexity to improve depth and reliability.

The Gems themselves could also be refined further. Some instructions are unnecessarily detailed, while other areas still lack clarity. Certain functions, such as generating Python code, may also be unnecessary depending on the purpose of the notebook.

As the system grows, continuous refinement will remain part of the process.

Conclusion

Building a Second Brain is something I’ve wanted to do for a long time.

I consume large amounts of information daily through books, videos, articles, and online platforms. However, much of that knowledge was never applied effectively. Over time, many useful ideas faded simply because they were never revisited properly.

This project helped me solve part of that problem.

More than just a portfolio project, this became a system I genuinely intend to continue improving and using in daily life. Instead of relying entirely on memory, I now have a structured environment where useful knowledge can be stored, recalled, and connected more effectively.

At the same time, the challenges remain.

Maintaining an effective Second Brain still requires continuous learning, reliable resources, and proper judgment. Without regular updates and careful curation, problems such as outdated information and inaccurate outputs can still occur.

This reinforced one important realization for me:

AI is only as good as we are

Artificial intelligence can organize information, summarize ideas, and accelerate learning. However, it cannot replace human curiosity, judgment, or understanding. The responsibility to think critically and continue learning still belongs to the user.

In many ways, building a Second Brain does not reduce the need for learning. Instead, it raises the standard.

That is the true function of a Second Brain.