Fake / Real: link fact-checker with Jev as judge

# Fake / Real: made with Jev

[Fake / Real](https://www.fake-real.live) is an English and Romanian evidence-review site for news, posts and claims. You paste a public link, and it returns an explained report: the factual claims it found, the outside sources it read, and a REAL, FAKE or not-enough-evidence result with the reasons.

This repository shows how the site uses [TypeSafe Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev), TypeSafe's System One model. The full application is closed source. The files here are the production Jev client and claim check, published so the integration can be read and run on its own.

## Watch how it works

[](media/fake-real-walkthrough.mp4)

**[Watch or download the full video](media/fake-real-walkthrough.mp4)** · 49 seconds · 720p · English narration and on-screen explanations.

See link entry, the live scan timer, JEV's evidence-checking role, a saved REAL example, a saved FAKE example, and the linked sources. The live documentation scan shown ended without a verdict; its credit was returned. The coffee reports are clearly labeled saved examples, with their original analysis times.

[Capture notes and editable video source](docs/video-source/PROVENANCE.md). Re-render from `docs/video-source` using `npm run render` (Node 22+, FFmpeg, and the pinned HyperFrames CLI).

## Where Jev sits in the pipeline

1. Firecrawl retrieves the submitted page. Gemini extracts at most five short, atomic factual claims.
2. Tavily searches for outside sources. Jev decides which search results are relevant enough to read (Noul questions, batched).
3. For each claim, Jev reads the cited excerpts and picks one of `supported`, `contradicted` or `unknown` (one Choice question per claim).
4. Jev also picks which claim is the article's main