The Oracle

Gameplay
Avatar
Oracle (Spotify Wrapped)

Inspiration

For inspiration we took a study by Joon Song Park, Generative Agent Simulations of 1000 people. in which we figured that AI agents with enough time and data can create accurate predictions onto peoples lives.

What it can do

The Oracle first takes in data from the user, either Resume input or questioaire. The ai agent then creates predictions based on deterministic data that we set up which allows for more accurate predictions

How we built it / Stack

We built our preject with a React frontend and Supabase backend, whilst most computation and comunication was done with typescript and javascript, whilst HTML/CSS for webapp.

Challenges we ran into

Before heavy constraints were put on the model, the model would create outputs that didnt match exactly what we wanted, but with testing and prompting the ouput is fine. Also user verification with the login page was trick with use of a database.

Accomplishment that we're proud of

Were very proud of The Oracle, it has an extensive interactive frontend that is user friendly for all ages groups ect. While the main function is even better with how many of the decisions you're given are accurate to whats given, being able to give custom inputs gives unlimited positiblies

What we learned

Through this hackathon we learned about how AI models work with data and the memory stored. Using LLMs for inference and behavioral prediction and deterministic formulas for node/action deltas.

What's next for Oracle

The Oracle could be expanded upon in many different ways, since its only a yearly cycle we can add tuning to be able to scale it monthly, weekly, or daily, which will give more accuracy to our output. We also can expand to career only simulations or find a way to make the metrics even more intricate and relevant for each and ever user :)