I talked to my book using ChatGPT and it changed how I read

Reading is one of my favorite things to do, but the biggest problem I have with reading is retaining the information I read. When I stumbled upon an interesting idea, I highlighted and took notes on it, but even then, I often found myself forgetting what I read soon after finishing the book. There are two main challenges to retaining information in a book in my opinion:

• Recalling ideas: Whenever I want to revisit an idea, I have to remember exactly where it is in the book. I can search the book (digital version required), but it can be time-consuming and even frustrating because I need to use very specific wording to find relevant results.

• Applying ideas: Many books have ideas and step-by-step frameworks that I can apply to my personal life, but the actual application can be overwhelming and labor-intensive, especially when the frameworks are complex.

So I thought: wouldn't it be cool if I could create a custom version of ChatGPT based on a book that I’m reading, so I can ask questions and interact with it? This way, I can learn and retain information in the book much better while making the entire reading experience more interactive.

Building the GPT

?? Step 1: Set up the knowledge base

I decided to pick The Millionaire Fastlane by MJ DeMarco to build a custom GPT on. It is a book I read recently and really liked, since it talked about topics that I really resonated with (entrepreneurship, wealth creation, financial mindset), and more importantly, it has many practical frameworks, which makes it the perfect subject for this project.

First, I gave the GPT the pdf version of the book that I owned as a knowledge base for it to reference (along with other data sources it already has access to). This will allow the GPT to draw on specific ideas and frameworks in the book when it needs to. It will also allow the GPT to emulate the style and tone of the author, something I’ll ask it to do in the next step.

?? Step 2: Specify the role, context, and style

Next, I give the GPT instructions about its role and some high-level guidelines that it should follow when responding to users:

Use the pdf version of the book uploaded under Knowledge as the main source of truth.

When the user asks a general question related to any concepts and topics mentioned in the book, give an answer based on what was written in the book.

When the user asks for help on a specific task, reference strategies and frameworks in the book to help users.

I also specify some guidelines about the style and tone that I want it to answer in. In this example, I want the GPT to answer in the same tone and style of the author, so I actually asked ChatGPT to describe the style and tone of the author for me, and use that information to form the instruction. Using the GPT to fine-tune the GPT!

Answer questions directly and assertively, in a casual and conversational fashion, like talking to a friend over coffee.

?? Step 3: Specify the protocols

Here is when the fun really begins. In business books like this one and many others, there are often step-by-step frameworks that the reader can follow to accomplish a certain task. For The Millionaire Fastlane, there are frameworks to:

• Assess business opportunity
• Make business decisions based on various criteria
• Calculate financial independence targets

Now what I can instruct the GPT to do now is to reference certain frameworks in the book whenever the user needs to accomplish a task related to that framework, then help guide users through that process step by step, in a way that is highly interactive and personalized.

If the user asks you to assess a business idea or opportunity, run through this process:

1. Assess the given opportunity using the CENTS framework in Part 7.

2. Format your answer in a table: each column is a factor in the CENTS framework, each row is the assessment of the opportunity based on the factor.

3. Give a final verdict on whether this is a good business idea, and also any suggestions you have on how to mitigate potential risks or shortcomings of the opportunity.

As you can see, I even specified to the GPT how I want it to format the answer in a certain way to make it easier to consume. Finally, I even ask it to offer suggestions on how to mitigate potential issues with the business based on principles mentioned in the book. From that point on, I can continue to go back-and-forth with the GPT about this idea.

What’s great about this is that the GPT can not only follow step-by-step instructions, it can also research information and generate original ideas related to certain steps. Here’s an example of this happening when I ask it to go through a decision making framework outlined in the book (e.g. step 2 and 4).

If the user asks you to make a decision between various choices, run through this process:

1. Ask the user for the general context of the decision, and the choices being considered.

2. Then, only when the context is given, brainstorm important factors that would impact the decision based on the context, then ask user to modify the list if they wish.

3. Ask the user to specify the importance of each factor (from 1 to 10).

4. Assess each choice based on all the factors based on your own research and reasoning, then ask the user to confirm your assessment before continuing.

5. Format your answer in a table: each column is a choice, each row is a factor.

6. Calculate weighted score for each choice using the WADM framework in Part 6, and make a conclusion about which choice the user should make.

There are two benefits to this: 1. It helps provide information and perspectives that I might not have thought of otherwise 2. It can act as a partner for me to bounce ideas off. This ability to go back and forth had really enhanced the reading and learning experience for me: It makes me feel like I’m talking with the author himself, especially with the instruction about style and tone I provided earlier.

?? Ready, Set, Action!

But it gets even better. One of custom GPT’s most powerful capability, actions, allows it to perform tasks instead of just writing, by talking to other applications via APIs. For this experiment, I thought it would be quite nice to connect the GPT to my personal Notion, so that as I talk to it, if there are thoughts or ideas that I want to note down, I can tell it go ahead and save it to my Notion, without me having to manually copy/paste the content. To do this, I needed to use Make.com, an automation platform that enables different tools to connect with each other and form a workflow. This step requires some coding knowledge (such as understanding of API), but it is by no means difficult.

With this action set up, I can simply tell the GPT to save notes to Notion, and it would do so without me having to go to Notion. Pretty handy!

The possibilities

Building this custom GPT was a fascinating experience. It's not perfect—about 8 out of 10 answers are spot on, but there are still some misses. Yet, having used the GPT in the last few weeks, I truly think that the potential for this is immense. The ability to interact with any materials in a personalized manner on demand makes GPTs like these a powerful tool for knowledge-based industries like publishing, coaching, and consulting. Imagine a future where every book comes with a custom GPT to enhance readers' experiences. Authors, coaches, consultants, and even YouTubers can have a GPT version of themselves, providing a new, interactive way to engage with and monetize their audience.

What do you think? Any ideas on how I can make this even better? Let me know down below ??

P/s: If you're curious and want to give it a try, here you go! The Millionaire Fastlane GPT