# FactDrop

> An AI-powered fact-checking engine that turns a claim into a sourced verdict in seconds.

_2026-03-22_

[Live](https://factdrop.org/) · [Source](https://github.com/sayeedjoy)

FactDrop grew out of the work at Rumor Scanner: human fact-checkers are
accurate but slow, and the internet ships misinformation faster than any team
can keep up. FactDrop is the first-pass triage — it gathers evidence, weighs
sources, and hands reviewers a head start instead of a blank page.

## What it does

- **Claim ingestion** — paste a claim, a URL, or a social post and FactDrop
  extracts the verifiable assertions inside it.
- **Evidence retrieval** — it searches a curated index of trusted sources and
  ranks passages by relevance and credibility.
- **Explained verdicts** — every verdict ships with the citations behind it, so
  a human can audit the reasoning rather than trust a black box.

## How it's built

A FastAPI service orchestrates retrieval and the model calls; Postgres stores
claims, sources, and the audit trail. The reviewer-facing dashboard is a Next.js
app, and the whole thing ships as containers so it runs the same locally and in
production.

It doesn't replace fact-checkers — it makes each one faster.