marzapower

For students

Arianna

Study from a manual built for you, not from scattered slides.

It takes a university course's material, even incomplete or badly written, and rebuilds it into an ordered manual, with exam questions woven in section by section. An assistant answers by citing the exact point in the manual, and quizzes tell you how much you really know about each part.

Open AriannaHow it started
A page from the manual Arianna rebuilt: chapter 4, Thermodynamics and enzyme kinetics, a seven-minute read, with the first paragraph highlighted.
A page from an Applied Biology manual, rebuilt from the course's slides.
Role
Sole founder. Designed and built end to end.
Status
Evolving. Web reader live, macOS app for building manuals, mobile reader in development.
Built with
Rust and Python for the builder, Ruby on Rails with Inertia for the web, React Native for mobile.
Updated
September 2026

Where it started

Arianna started as a tool for myself. When I began studying psychology, the course material was what it usually is at university: slides in PDF or PowerPoint, handed out by professors, often badly written, disorganized, incomplete, sometimes wrong or simply out of date.

On top of that, courses hand out two disconnected things: a set of study material and a separate set of example questions, which may have little to do with the lessons. Nothing ties them together, and there's no way to know how much of what the exam will ask is actually covered.

So instead of studying from the raw slides, I built the tool that rebuilds them: it takes everything the course provides and turns it into a single manual to study from. The macOS app I use to build the manuals is the same one I use myself, every week.

What it does for students

Six things, in the order you meet them.

  1. 1

    Rebuilds the manual from the course's material

    It ingests slides, handouts and exam questions and builds a brand-new manual from them, ordered from a teaching standpoint: topics arrive in the right sequence and at the right level of depth. A reference course runs from 24 to over 70 lessons.

  2. 2

    Weaves exam questions into the text

    While it builds the manual's structure, it attaches to every important section the questions that relate to it. The question set and the study material, which the course hands out separately, become one thing. It also runs randomized quizzes, drawn from the source questions and generated on the fly by its own AI.

  3. 3

    Answers by citing the manual

    A tutor tied to the manual clarifies doubts about the material with direct citations to the sections they come from. An answer always traces back to where the manual actually says it, not to what the model remembers.

  4. 4

    Tells you how much you know, chapter by chapter

    A competence score per chapter, built from your correct and incorrect quiz answers, shows where you're strong and where another pass is needed. That way you can focus your review on the topics that need it.

  5. 5

    Helps you pick up the thread and organize review

    Every section ends with a gesture: to review, completed, next. The manual remembers where you were and what you left open, so review starts there instead of from page one. Based on your reading pace, it also estimates when you'll finish the manual.

  6. 6

    Builds a manual even without starting material

    From the macOS app you can also pick a topic and let the AI build the manual, with no certified slides or handouts at all. It's the most open mode, for studying something no course has handed you yet.

How it's built

For anyone who wants to look underneath.

I chose to rebuild the material into a structured manual, instead of just indexing the original slides: its sections become the common reference for exam questions, the tutor's citations and the competence score. That lets a student move from reading to checking their knowledge while staying on the same topic.

The tutor answers only from the manual's content, never from memory, and every answer carries the citation to the section it comes from — so a student can always verify where the manual actually says it, instead of taking it on faith. The same manual also feeds the quizzes and the competence score, and is read by the macOS app I use to build it, by the web reader, and soon by the mobile reader.

Where it stands

The web reader is live at arianna.marzapower.com. The macOS app is the tool I use to build the manuals. The mobile reader is in development. The next feature planned is flashcards, generated from each chapter's key points.

Evolving, September 2026

Building a tool for people who learn?

I made every decision on Arianna alone, from the architecture to who uses it. If you're making yours, let's talk.

Get in touch

More work

  • Fabulous FactoryFoundations already in place for your AI agent. It builds the product.
  • EmoviaBringing AI into clinical practice, starting with the time between sessions.
  • Power PuzzlesPuzzles you solve by reasoning, never by guessing.