Cosine similarity fails because of Entity Drift. The viral claim morphed between "Somalians" and "Haitians", and between "cats/dogs" and "park ducks". A purely mathematical vector comparison scores these too low to match. The system must use Structural Schema Extraction to map: [Immigrant Entity] → [Consuming] → [Domestic/Local Animal] → [US Location].
Mainstream indices (Common Crawl, Google) do not index Telegram, 4chan, or closed Facebook groups. The system utilizes specific dark-channel ingests (e.g., Telethon API for Telegram, 4plebs for 4chan) to continue the Time-Zero traversal once the claim vanishes from the surface web.
The claim is stated as fact on national television to tens of millions of viewers. Standard verification systems assign this a high trust score due to the speaker's authority and platform.
Accounts with 2M+ followers post AI-generated memes and state the claim as breaking news. The algorithm pushes this into mainstream consciousness.
The system loses the claim on the surface web. It switches to Dark Channel indices. It locates the claim heavily amplified by extremist Telegram channels using the exact structural schema, coordinating a push to surface-web platforms (Gab/X).
The absolute origin. A fourth-hand anecdote posted by a local resident ("My neighbor's daughter's friend..."). The post was screenshot, removed from Facebook, but preserved on a 4chan imageboard archive.