Skutor

• 

  App interface in laptop browser

• 

  App interface in iPad browser

Inspiration

AI assistants have transformed learning, but have been observed to negatively impact students' reasoning abilities. Over-dependence on this assistance causes students to shortcut the learning process. We set out to create a learning interface to leverage AI agents, without sacrificing the student's opportunity to develop problem-solving skills.

What it does

We integrate AI assistant into a page interactive, digital notebook. Our assistant analyzes and processes this handwritten content, providing the user with step-by-step instruction to work through their problem. This allows the user to utilize AI assistance to enhance learning without sacrificing developing a crutch.

How we built it

Using React and Typescript for the UI, with a Typescript backend, and OpenAI stream integrations.

Challenges we ran into

A major challenge was engineering an effective prompt to always guide the user on the next step. Creating a stable user interface that can be interacted with via Apple pencil.

Accomplishments that we're proud of

We are proud of how seamless and comfortable the interface seems. We are also proud of how well we were able to develop the reasoning and guidance abilities in the assistant.

What we learned

We learned how to develop a robust backend for secure streaming. We also learned how to optimize image reasoning models to provide effective feedback.

What's next for Skutor

We plan on further improving Skutor by incorporating more options for which model to use, alongside allowing users to copy and paste screenshots into Skutor to make our app more convenient. Additionally we plan on incorporating a text box to allow users to type out their problems.

Built With

• openai
• react
• typescript