Recently, I have dedicated more time to sharpening my knowledge in my profession as a product manager. This means studying the latest best practices in product management. When I read, I thoroughly highlight key insights in the text using Readwise.
Effectiveness of the quiz
Simply revisiting highlights is helpful, but I've found that quizzing myself is the best way to truly retain knowledge from what I've read.
Creating flashcards, as we did in school, is one effective strategy here. However, in the age of AI and natural language processing, an even better solution may exist.
Using Claude AI to Generate Quizzes
Claude AI by Anthropic is an AI chatbot, similar to ChatGPT, with a text-based interface for leveraging AI. I primarily use Claude because of its file upload feature and approach to AI training.
Tools like Claude AI and ChatGPT excel at transforming unstructured text into structured outputs based on provided instructions, called prompts. Here, I can take book highlights and transform them into a multiple choice quiz format.
I downloaded my highlights from Readwise as a Markdown file containing the book title and all highlights. In Claude's interface, I entered the following prompt explaining how to interpret the highlights and convert them into a multiple choice quiz. I also uploaded the Markdown file.
Format the attached markdown file content into a multiple-choice quiz.
Use the header text as the title and take the remaining text below that text to generate a question, and a list of answers, three false answers and one correct answer all in randomized order. Make sure to note which answer is the correct answer for each question.
After submitting the prompt and file, Claude returned a response transforming the highlights into a quiz. I can use this text directly for flashcards or to create an interactive Google Forms quiz.
While a human could certainly manually create questions, answers, and a quiz from highlights, Claude's AI enables this instantly, saving valuable time. With Claude handling the busy work, I can devote more time to actually studying the content.
Advanced Prompting for Custom formats
Beyond simple flashcards, I wanted to try something even more advanced. Years ago, I built a basic quiz framework that uses a configuration file to dynamically generate quizzes. My idea was to make the code reusable by varying the quiz content based solely on the config file.
Just as Claude previously transformed my highlights into a multiple choice quiz, I could now use the following prompt it to output highlights in the code structure my framework requires.
Format the attached markdown file to the example multiple choice quiz json.
Use the header text as the title as labeled in the example and take the remaining text below that text to generate a question , and an array of answers, 3 false answers and one correct answer all in randomized order.
The index number of the correct answer should be noted in the answer object (as noted in the example) as well as an explanation to that answer.
Each text block below the header, separated by a newline represents a new question.
Final outbut should match the format below.
"title": "title here",
"highlightcolor": "color here",
"questions": [
{
"question": "What type of animal is a seahorse?",
"answer": {
"number": 2,
"explanation": "The seahorse is a bony fish. Their bones are located on the outside of their bodies."
},
"choices": [
"Crustacean",
"Arachnid",
"Fish",
"Shell"
]
}
]
With some refinement over multiple tries, Claude successfully interpreted the same highlights file and output content matching my custom configuration format. I simply copied and pasted this into my framework's config file, immediately generating a full quiz.
Summary
This experiment provided valuable insights into prompting complexity levels and output examples for steering AI. As AI capabilities grow, exploring creative applications like this becomes increasingly valuable.
In summary, this showcases approaches for repurposing AI tools like Claude for personalized uses beyond their core functionality. Converting highlights into quizzes is one example, but the possibilities are endless. I look forward to continuing to explore imaginative ways to leverage AI for productivity.