What it is
A Manifest V3 Chrome extension for X (x.com) tweet detail pages that collapses spam, bait, off-topic and AI-filler replies. Two-stage: free local rules in rules.js, then Jev for the rest. Users see hidden replies as collapsible grey bars and learn from manual corrections.
How it uses Jev
For replies passing local rules, background.js batches 8 replies and sends four noul questions to TypeSafe jev via OpenRouter alpha/decisions: promo/spam, engagement bait, off-topic, AI filler. Any score ≥ threshold (default 0.75) collapses the reply into a one-line 'hidden · category · probability · click to expand' bar. Verdicts are cached in chrome.storage.local by tweet id and by author+text hash to avoid re-billing.
Primitives:noul
Technique worth stealing
Two-key verdict cache (tweet id plus author+text hash) so repeated content and id changes never re-query jev.
Try it
Load unpacked in chrome://extensions, open an x.com/ /status/ page. Free tier or supply an OpenRouter key.
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.78
- Shows a System One patternyes0.88
- Handles uncertaintyyes0.77
- Measuredno0.06
- Runnableyes0.98
- Worth recommendingunclear0.59
- Model replicano0.03
- Problem scopescore on a 0–2 scale1.21
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.