ClearLens

example of live agent history
source example
description of problem
title page
evidence graph
final claimlens report

I noticed that AI research tools can give wrong or misleading answers, even when they include sources. I wanted to build something that checks its own research instead of just giving an answer.

Untitled is an AI research agent that finds information, checks facts, and looks for evidence that disagrees with its own answers. It also detects when multiple websites get their information from the same source. Users can explore everything through an interactive graph.

I used the Pi Agent as the backend. I created a research system that searches the web, organizes evidence, and generates reports. I also built a website that displays claims, sources, and their connections.

One of my biggest challenges was making sure the AI connected the right evidence to each claim. I also had to fix problems with detecting conflicting sources, running follow-up searches, and getting the website to work smoothly.

I'm proud of building a working research agent that questions its own results. My system can find conflicting evidence, track where information comes from, and automatically research weak claims. I also built an interactive graph and tested the full application.

I learned that having a lot of sources doesn't always mean information is reliable. I also learned how to coordinate multiple AI agents, build evidence graphs, and test an AI system using real research questions.

I want to improve the accuracy of my research, make the website easier to use, and support more types of sources. Eventually, I hope to turn Untitled into a tool that helps students, researchers, and everyday people find information they can trust.