Inspiration
Wanted to participate in a live hackathon to learn more about AGI.
What it does
Attempts to gain a good score in the 11 synthetic web site misc. tasks
How we built it
Use Claude Code Sonnet 3.7 + Hierarchical Planning + Native Accessible Trees
Challenges we ran into
It was difficult to really understand how to make this process better. Also, I needed to climb the learning curve on how this all works -- I'm a newbie!
Accomplishments that we're proud of
Got it to work but initially only concentrated on one site. Later, tried to get make the ai agent more general and work on more than one site, only for the easy tasks.
What we learned
Learned tons of stuff, and about how Silicon Valley benchmark's web tasks. It was interesting that even the simplest human tasks are difficult for LLMs to do well.
What's next for MyFirstHumbleAGIHackathonSubmission
Watch the pros do this live!
Log in or sign up for Devpost to join the conversation.