Video unavailable. Open the source below.
NewX Scanner
Label visible X posts with typed Jev judgments about promotion, engagement prompts and information density.
Added to Jevfast
How it uses Jev
x-scanner sends a post's extracted text to Jev with six typed questions in one request. The default set includes Noul checks such as whether the post ends with an engagement prompt and a Score question about how much specific, verifiable content it contains. The extension displays the answers and flags values that cross configured thresholds.
What Jev decides
Assess whether a post asks readers to engageExample answers · not a recorded Jev response · Source ↗
Question 1 · engagement bait
YOUR APP
INSTRUCTION
Does the post end by asking or nudging readers to interact, such as replying, reposting, liking, following, or bookmarking?
STATE
State includes a post's extracted text. The extension builds one request with its enabled dimensions; engagement_bait is one default Noul question. The illustrative answer below is not an observed Jev output.
JEV · NOUL
YesNo
Score the post's information densityExample answers · not a recorded Jev response · Source ↗
Question 1 · info density
YOUR APP
INSTRUCTION
Rate the amount of specific, verifiable content on the source-defined four-level scale.
STATE
The same post text is sent with the enabled dimensions. info_density is a Score question with four ordered levels from no checkable details to dense, specific content. The example has no captured score value.
JEV · SCORE
0
App workflow
- Detect posts near the viewport
Mutation and intersection observers find X posts and queue them when they approach the viewport; the queue can drop posts that scroll away before processing.
- Extract and skip ineligible text
The content script extracts post text and metadata, and skips promoted posts or posts with no text.
- Build and send one request
The service worker builds the enabled typed questions and sends them with the post text to Jev, applying bounded concurrency and retry rules.
- Cache and display results
Answers are cached by post id, thresholds determine which labels are flagged, and the chip/card shows values, usage, cost, and latency.
Why it is interesting
It evaluates visible posts near the viewport as the user scrolls, then caches answers by post id. Typed Noul and Score outputs support different kinds of behavioral labels while the extension tracks cost and skips promoted posts and posts without text.