Skip to content

michaelfromorg/tutor

Repository files navigation

Tutor

A screenshot of the app.

An experiment with AI and whiteboards.

Tutor is an infinite canvas intended for students to complete homework problems. Each page is a question, and the AI is able to either chat or draw to help the student solve the problem.

For example, if you're given a basic problem like...

A school auditorium has 18 rows with 32 seats in each row. If tickets cost $12 each, how much money would the school collect if every seat is sold?

The large language model can...

  • Provide you hints
  • Draw pictures to help you visualize the problem
  • Correct your mistakes

Notes for judges

  • Relies on a mix of an infinite canvas and chat UI (that can interact with the canvas!)
  • Entirely vibe-coded, mostly from 1pm; using Windsurf with a mix of o3, sonnet-3.7, and sonnet-4
  • Persist user questions to local storage, and remembers past chats for prompt re-use
  • Basic visuals that match the tldraw style

Future ideas...

  • Students upload their homework PDF, and it's immediately broken into a series of question pages
  • AI nudges students towards the correct answer as they work, catches mistakes as you go (instead of chat -- chat should be secondary)

About

An experiment in AI and HTML canvases. Using tldraw.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages