Poll Master
Deep Research with human feedback through proof-of-human polling
Project Description
Fake content and bot-generated opinions dominate many of today’s online platforms, distorting public discourse and making tools like Deep Research unreliable for subjective or easily manipulated topics.
This breaks a key use case: large-scale public opinion research. Without trustworthy data, users are forced to rely on slow, expensive in-person methods.
Our solution combines Deep Research with verified community polling using WorldID. Polls are created automatically for any research topic prone to manipulation, and only verified humans can vote—ensuring results reflect real people, not bots.
This is a system where research isn’t just automated—it’s human-backed, trustworthy, and economically incentivized.
How it's Made
The project is built as a World MiniApp with a React frontend and connects to a local GPT-4-powered Deep Research instance that autonomously generates insights and polls. It features an Express backend and a MongoDB database. Identity and payment verifications are handled via the World SDK and validated on the backend. The Deep Research instance has access to all poll data and leverages RAG capabilities through langchain-open-deep-research.
