TutorAI-written personalized books that learn how you learn

Personalize

You can't tutor a reader you've never met, so after entering your OpenAI, Anthropic, or Gemini keys, the app first has you fill out your learning profile. To save time, an AI-led interview can gather all the answers with a series of simple questions.

Title and description

Every book starts with a title and a description. Type your own, or let the AI propose both.

If you don't know what to put, you can ask the AI to expand your existing skills, learn a new complementary skill, or explore a completely new topic and it'll choose an appropriate title and description for you.

A topic, optional details, and a length. Suggest proposes the book for you; everything stays editable before generation.

Table of contents

From that brief, Tutor first just drafts a full table of contents. This way you can make sure the book aligns with your goals before any chapters are written.

If you're not happy with the table of contents, you can chat with the AI to make any necessary adjustments.

First chapter

Approve the outline and Tutor writes the first chapter. Just one. The rest of the book stays intentionally unwritten.

Chapters run about 1,500 words, so they only require about five to ten minutes of reading.¹ Short enough to binge read multiple chapters, long enough that each chapter could cover its subject matter in sufficient depth.

Chat

Select any passage and an inline chat opens, anchored to that exact text, with the context of the entire chapter and the subject matter of the book. When you ask "what does this mean," the model already knows what "this" is.

When you're done with the conversation, simply close the chat and you can continue reading right where you left off. This feature of being able to zoom in, then zoom back out is a major improvement over the experience of learning with a simple LLM chat bot.

Highlight a sentence, ask a question, start a conversation. The AI has context on the chapter and the selected text.

Rich content

Chapters are more than prose. Math renders as real equations, code arrives syntax-highlighted, and ideas easier seen than said become Mermaid diagrams.

Callouts lift the one thing you shouldn't miss out of the paragraph and onto the page.

KaTeX math, highlighted code, and callouts, generated inline wherever the subject calls for them. You can also customize the reading mode settings to suit your preferences.

Chapter quiz

Every chapter ends with a short, skippable, multiple-choice quiz on what you just read. The quiz serves two purposes. First it helps you reflect on what you just read to aid retention. Second, it provides feedback to the AI to tailor the next chapter to your needs.

Feedback

After each chapter you can say what worked and what didn't, or just skip ahead. This info will be used to generate all subsequent chapters, as well as the suggested updates to your learning profile when you finish the book.

Plain-language feedback on the chapter you just finished, captured while it's still fresh.

Next chapter

The next chapter is written only after the last one is read, generated from your quiz results, your feedback, your learning style, and the structure set forth by the table of contents.

End-of-book quiz

At the end of the book, there's a final quiz that tests whether you can synthesize information across multiple chapters. After taking the quiz, you can write the book. So you remember for future use, which books you like most.

Learning profile

While Tutor was generating each chapter, it was also creating a list of skills associated with each chapter. Once you finish the book, Tutor uses those skills as well as your feedback along the way to give you suggestions on how to update your learning profile.

Tutor can propose adding new skills, update your existing skills to a new competency level, update your learning profile's free-text field, or update any of your discrete preferences such as whether you like metaphors or humor in your books.

All of these proposed changes are easy to view, approve, or override. So the user is always in the loop and always in control, while the AI handles the busy work.

Series

You can group related books into a series for easy categorization.

Organize

Powerful sort and filter tools help you find the right book at the right time. You can use Gemini to create a unique personalized cover image for each book.

Audiobook

Generate a narrated version of any book, with chapter markers and a voice you choose. The audiobook generation happens completely offline and for free using an open source model, with many voices to choose from.

Import and export

Export any book as an EPUB to read in your e-reader of choice, or import one someone else generated and make it yours. Nothing you make is trapped inside the app.

Open and yours

Tutor is GPLv3, BYOK, and totally free to use. Point it at OpenAI, Anthropic, a local model, whatever you prefer, and the app uses your keys directly with no server in the middle.

I'm not the bottleneck or the billing department for someone else's reading habit. The app outlives any one model, any one provider, and me.

Offline by design

Every book is plain markdown on your disk. No database, no cloud, no account. The library is a folder stored on your computer, so you can back it up, share it, or do whatever you like with it.

Advanced generation via MCP

When generating a book, Tutor has an "Advanced" mode where you can copy/paste a prompt into your terminal that will generate your book for use using Claude Code. Behind the scenes, Claude Code will use a specialized MCP and skills.

This mode is helpful if you want enhanced coherence, if you want to leverage any MCPs or skills you already have, if you need an agent to search the web for you, and if you want to utilize more sophisticated generation techniques such as "plan mode."

Reflections

I believe one of the biggest bottlenecks in the AI-transformation of the world is the speed at which humans can learn and adapt to new information. I've read dozens of books with Tutor. It's helped me learn new information faster while keeping it engaging and immediately useful for my specific goals.

Tutor is a great sandbox for me to experiment with an AI-first product that I have full control over, where AI is the substrate of the app that touches nearly every feature, instead of something tacked onto an existing app

Download tutor today at https://github.com/rsml/tutor/releases or learn more at rossmiller.dev/tutor.