Schedule

Deliverables

1. Any files you would like to upload including valuable information about your project
2. Any links/urls you would like to include (working demo, video, github page, figma share)
3. Contact email address you check regularly
4. Country location (This will help us determine eligibility for shipping prizes)
5. Please also include a description of your project & the issue(s) it helps solve.

Ideas

• conversation coach app: streamlit_app · Streamlit (convo-hackathon.fly.dev)
• personalized education/events → learn what you like and dont like

Inspiration

• Learning in our experience strongly depends on personal motivation and the connectedness of information, did you ever go down one of those Wikipedia rabbit holes?
• You read a text and want to be sure you got the gist of it, like reading a Wikipedia article or some tutorial

What it does

• Challenges you understanding of studied content and helps you memorize better and faster
• After reading an informative test, Qbot asks you question about the content in form of quiz.
• You can answer the questions and earn point on correct answered questions, thus improving your score.
• The better your score, the better your understanding of the text

How we built it

• Prompt for GPT-3 with custom prompt and the article as context
• Browser extensions to integrate seamlessly into the user flow

Challenges we ran into

• Validating the answers will require further work, possibly the collection of a custom dataset

Accomplishments that we're proud of

• Beneficial use of generative AI
• Improving the learning journey of our users

What we learned

• Prompt engineering does not guarantee reliable results
• Time zone conversions are tricky

What's next for Qbot

• Build an appealing interface
• Fine-tune the given questions and the scoring
• Launch on Chrome’s and Firefox's extension marketplace
• Market the product on Social Media

Built With

• browser
• extension
• typescript