Inspiration

One of our team members was trying to stay updated on grad school application deadlines and requirements. Their first thought was, “I should build a centralized newsletter or website for aggregating all this information!” Then they realized they had an AI hackathon the next day and thought, “I should use AI to solve this problem for all people who want to stay subscribed to cool and relevant links on the Internet!” And here we are.

What it does

Our software takes in a list of links, topics (and in the future a persona) and generates a newsletter with recent and relevant content.

How we built it

AI! We basically had to rebuild a version of Google’s web crawlers and use AI to semantically determine the relevancy and recency of the site content.

Challenges we ran into

The newsletter generation functionality works incredibly well in ChatGPT. But it underperforms through the API due to context window sizes (the API models are not cutting-edge). So we had to classically scrape by parsing out the HTML which led to semantic information loss. This should be solved relatively soon!

Accomplishments that we're proud of

Definitely the classical scraping and working around the context window limits, that was a doozy. We also used Clearbit on the frontend to get the website logos and other relevant information which we think was pretty resourceful.

What we learned

Search is hard. AI is powerful.

What's next for Quirk

We want to add newsletter/subscription version control. We want to see what our heroes are reading on a daily/weekly basis, show others what we’re interested in, and subscribe to the same links our heroes do! This 1) allows for classic network effects and 2) incentivizes quirkiness because nobody wants to be seen as reading all the basic stuff that everyone else is reading. We hope that this will preserve some of the cool parts of the Internet that haven’t been discovered, in much the same way that people take pride in finding great, but relatively unknown, artists.

Built With

Share this project:

Updates