How I Study With AI

Introduction

#

This post describes how I use several AI tools to help me study topics of any kind.

For this, I usually go through the following 3 steps:

1. Assemble the learning material with GitHub Copilot
2. Create cards with AI Studio
3. Create audio summary with NotebookLM and AI Studio

Assemble Learning Material

#

First, I collect all the study material I already have. For topics where I don’t have any material or just need a lot more, I use an agentic tool like GitHub Copilot, which I am most comfortable with as a software engineer. Also, running the agent locally has the benefit that you can easily split your material across multiple files instead of bundling it in a large response. In the future, however, I should also try this approach with other agentic tools like ChatGPT Deep Research.

I then pass the agent a prompt similar to this:

Based on the topics and subtopics represented as topics below, for each bullet, create a comprehensive section, explaining all related concepts in detail. Take your information from <some-related-website> and all sub pages. Create a markdown file for each topic. Place the markdown files in a <name-of-the-subject> folder.

This leaves you with an organized number of markdown files which can already be used for studying by themselves.

Create Cards

#

Next, we use the markdown files to create flashcards for the Mochi app, which lets you easily import cards for studying using spaced repetition.

This can be achieved with the same local agent as above using this prompt:

For each topic markdown file, create flashcards in csv format, where the first column is the question and the second column the answer. Do not use headers.

These CSV files can then be imported in Mochi using the Import menu and the CSV option. Make sure to set Quote to ".

Create Audio Summary

#

Create a new notebook for each topic in NotebookLM and either upload the markdown file or copy its content via Paste text > Copied text. Then, click on Audio Overview to create a podcast based on the imported study material.

While the interaction of the two virtual podcast hosts can be entertaining, I often prefer a more concise single-speaker audio summary, which we can obtain using the following steps (inspired by Tina Huang):

• Download the podcast audio file in the playback section of NotebookLM.
• Import the audio file in a tool like AI Studio.
• Transcribe it using a prompt like Accurately transcribe the uploaded podcast with two speakers. Do not use timestamps.
• Convert this transcript into a single host format and remove all extraneous information and commentary. Stay concise but do not remove any facts. Write it such that it can be read by a single speaker.
• Copy the output.
• Go to Generate Media > Generate speech and paste the output from the previous step.
• Wait until the generation has finished and either listen in AI Studio or download the audio to listen offline.