How I Fixed a Broken RevenueCat Subscription in a Day with NotebookLM

Last Tuesday, I was chilling when I got a message from a client I had worked with previously. He had built an app, but the subscription was not working. He had been promoting it on different social media platforms for quite some time, and his app was starting to get installs.

Now, he was getting a lot of subscription failure alerts, which was not good for the app's reputation. We hopped on a call to see what was going on.

The app was built in Expo and used RevenueCat. I understood the problem and had to quickly find a solution for it.

Learning RevenueCat Fast (NotebookLM)

I have been using NotebookLM to get the summary of YouTube/PDF videos for quite some time. It saves a lot of time, as I can get the gist of the information instead of going through the whole resource.

I had to learn about RevenueCat fast as I only used StoreKit 2 on iOS for payments. I planned to use the same method to learn about RevenueCat—how it works, and its best practices.

I fed RevenueCat tutorial YouTube videos into NotebookLM along with its documentation. It helped me quickly understand how the whole process works, the code implementation, and how to handle different payment scenarios for an app.

This process might have taken 1–2 days with a traditional learning path, but I got the whole picture in a couple of hours.

The Fix

Once I got familiar with RevenueCat, I quickly ran an audit of the current subscription code with Cursor to find the missing links.

Now, I could understand the reason behind the subscription failure. The previous developer had missed the scenarios of subscription upgrades and restored purchases.

Finding the missing link was a crucial step, as I was able to create a plan in Cursor to fix it. Cursor helped me fix these issues one by one and run a simulation of each subscription scenario to see if there was still any missing link.

The fix was done within a day, and a new app version was pushed to the App Store. The app got approved the next day, and everything was fine.

The client sent an email to all the users who had missed the subscription renewal and got the subscription restored, helping them recover lost revenue and rebuild user trust.

My Simple Learning Framework

1. Curate Your Knowledge Sources

• Official documentation (PDFs, websites)
• Top 3–5 YouTube tutorials (paste URLs)
• Reddit threads where experts discuss real problems
• GitHub repositories with good README files
• Your own half-baked notes from previous attempts

Pro tip: Don't upload random, unrelated files. I once mixed JavaScript docs with Python tutorials in one notebook. NotebookLM got confused, and so did I. Keep notebooks focused on ONE topic.

2. Generate Study Materials

• Study guide → key concepts, definitions, glossary
• Mind map → visual overview of how everything connects
• FAQ → common questions answered from your sources
• Timeline (if applicable) → historical evolution or version changes

3. Let It Make Me a Podcast

If you don't like reading, you can generate an audio overview as well. Let two fake hosts argue about my material. I walk the dog and listen. Hearing concepts out loud locks them into my brain way better than silent reading.

4. Active Learning with Q&A

Once you understand the topic, you can start asking questions related to your problem.

Example: "I have a React Native app with RevenueCat integration. When I try to subscribe, the app throws a subscription cancelled error even though I gave the permission."

This will help you understand the topic better, and you can get better answers for your problem.

Common Mistakes to Avoid

Mistake #1: Treating NotebookLM Like a Magic Button

I tried to upload resources and then generate audio to learn a topic without verifying the docs or topic. It shared a sloppy understanding of the topic. You should treat it as a companion in learning instead of replacing the learning method itself.

Mistake #2: Uploading Too Many Sources

This goes without saying. You should only upload resources that you find helpful. Too many resources does not mean you will get better results. It's "garbage in, garbage out" when it comes to AI learning.

Mistake #3: Not Testing Your Understanding

Don't use NotebookLM to generate a learning guide and then move on. Always test what you have learned by generating a Q&A and verifying your input. It will help you get better at a skill, not just surface-level knowledge.

Power User Tips

Create a master notebook on a topic + specific topic notebook

You don't want to lose your knowledge on a major topic and go through the whole process again. It's better to keep a master notebook on a topic (React Native, marketing, vibe coding). If you want to learn a specific topic (RevenueCat), create a separate notebook.

Combine NotebookLM with ChatGPT

NotebookLM can help you generate a learning guide based on your topic or a step-by-step guide to fix a problem like I faced for RevenueCat. You can then use that guide to generate code, etc. with the help of ChatGPT.

Better feedback loop

Reading about a topic is not enough; you have to test your understanding from the perspective of an expert. Simulate an expert to generate Q&A to test your knowledge.

Example: "I have uploaded 10 topics on agents. Now act as a Senior ML Engineer and generate 10 Q&A to test my understanding of the agents."

Real-World Use Cases

There are plenty of ways in which you can use NotebookLM to make your life easier. Here are a couple of examples.

1. Learn a Topic

This is the most useful use case of NotebookLM. It's built to process large amounts of data from different resources and present that in a structured way.

Example: learning about coding, marketing, etc.

2. Interview Prep

Another great use case is interview prep. You can feed resources of the topics that you want to learn and then ask NotebookLM to simulate an expert on that topic and ask questions.

3. Build a Mentor

This is a very creative use case of NotebookLM that I have seen. If you follow any person online, you can build a mentor around their available knowledge.

Example: one of my friends, Arvindh, has created a mentor by feeding Jason Fried interviews and blogs in NotebookLM. It helps to get feedback from the perspective of that person.