S.L.O.P.

Inspiration

We struggled to find an original hackathon idea. Existing tools gave us ideas without much novelty, so we built S.L.O.P. (Similarity Lookup for Originality Predictions) to show what exists and help us find what’s different.

What it does

S.L.O.P. searches thousands of projects, companies and the web. A multi agent swarm debates similarities, verifies citations, and maps related projects as islands around your idea. An evidence-based score measures originality, while a conversational coach helps you improve your idea.

How we built it

We used Elasticsearch for hybrid search, Baseten for model inference, openJiuwen for multi-agent orchestration, and GPTZero for citation and writing checks. Exa and Browserbase support web research and FastAPI streams results to a Next.js and three.js interface.

Challenges We Ran Into

Short pitches confused rerankers, duplicate sources distorted our metrics, and models produced unreliable outputs. So we added relevance checks, entity resolution, and code-enforced verification and rebuilt our cluttered dashboard around an interactive map.

Accomplishments that We're Proud of

Making search results into an interactive island map that makes related ideas easy to explore.
Turning our own struggle to find an original hackathon idea into a tool other builders can use.