User repository on GitHub
- Forks
- 1
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Signal & Noise
A Chrome extension that hides posts you do not want on your X timeline: tap noise bubbles or type rules in plain words, and Jev judges each tweet one-shot before it scrolls into view. Good hide and Shouldn't have hidden answers stay local as an eval set.
- Stars
- 2
- Language
- HTML
- Reading time
- 3 min
- Published by
- Community
- Author
- icpmacdo
- Last verified
- Oct 2026
- Languages
- JavaScript
- Platforms
- Chrome extensionBrowser
- Capabilities
- Open sourceRequires API key
Highlights
- Each tweet gets one choice question — keep plus one label per mute — on pinned model jev-1.13.0; hidden when P(keep) is below the 35% default.
- Tweets stay invisible until judged; Jev answers in about 150–200 ms and posts show after 2 s anyway if it is slow or down.
- Install unpacked from chrome://extensions; pick TypeSafe Jev (default), OpenAI Decisions with gpt-6-luna, or a self-hosted /v1/systemone server.
- Tweet text, handle and image alt text go to the chosen model; keys stay in chrome.storage.local and always-show accounts are never sent.
- Good hide and Shouldn't have hidden answers are kept locally in feedbackLog, described as the start of a real-world eval set.
Watch out
Grounded in the full repo README and GitHub metadata read on 2026-10-15. The README gives no measured accuracy or precision for the hides — each verdict is a one-shot judgment cached per tweet id. The repository declares no license, and the os/ menu-bar app and decision pages are documented as experimental.
Questions about this entry
Short answers drawn from the linked project or page.
An open-source Chrome extension that hides posts you do not want on your X timeline. You tap common kinds of noise as bubbles — political rage bait, crypto shilling, engagement bait, spoilers — or type your own in plain words, and Jev judges each tweet against them before it scrolls into view.
Load it unpacked: open chrome://extensions, turn on Developer mode, choose Load unpacked and pick the folder. The settings page opens, where under Model you pick TypeSafe Jev (default), OpenAI Decisions or a local server, and paste its key; taps and changes save as you go. Then open x.com.
TypeSafe Jev is the default, sending one tweet per request to api.typesafe.ai/v1/systemone. OpenAI's Decisions API (gpt-6-luna) is the alternative and can also see up to two post images, so the Memes bubble only works there. You can also point it at any server speaking the /v1/systemone protocol.
Content code finds each tweet article as X renders it — deliberately ahead of your scroll — and holds it invisible until Jev answers, keeping its space so nothing jumps. Each tweet gets one choice question with keep plus one label per mute; it is hidden when P(keep) is below the strictness setting, 35% by default.
Tweet text, the author's handle and image alt text are sent to the model you chose for judging; on OpenAI with images on, the post's pictures are sent too. Nothing else is sent, keys stay in chrome.storage.local, and tweets from accounts on the always-show list are never sent at all.
Yes. After Show, a post asks Good hide or Shouldn't have hidden this, and those answers are kept locally in feedbackLog as the start of a real-world eval set. The always show list still wins, replies to a hidden post hide under the same bar, and the tweet you opened directly is never hidden.
Two experimental pieces ship alongside it: os/, a macOS menu-bar app that applies the same idea to everything on screen by judging pixels with OpenAI's Decisions API and blurring the matches, and decision pages built from synthetic tweets scored by Jev (Threshold Lab, Run Planner and Taste Test).
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: An X timeline labeler that flags engagement bait
X: Unclutter: an ad and slop blocker that runs on Jev
X: dmx.to reshapes the timeline in plain language
X: A real-time ad blocker that classifies every DOM element
More Jev resources like this
A Manifest V3 Chrome extension that sends each LinkedIn feed post's text to TypeSafe's Jev with a rubric of narrow yes/no and 0–2 questions, then blurs only the posts that plain code marks as filter — uncertain and failed checks stay visible, with one-click reveal.
Jev Tools & CLIs#community#catalog-import#madewithjev
Chrome extension that classifies X posts with Jev and blurs chosen expression categories — discrimination, provocation, hatred, calls to attack, cynicism, resignation — behind a translucent overlay you can reveal with one click.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
eijiaraki/toxic-filterJavaScript
eijiaraki/toxic-filterJavaScriptA Chrome extension (MV3) that scores every post entering your X timeline with a single Jev call — worth engaging plus selectable tags like spam, buzz, misread and AI-ish, shown as calibrated probabilities in the avatar column, with the goal you pick steering the worth-engaging question.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ddnim/jev-tweet-radarJavaScript
ddnim/jev-tweet-radarJavaScriptAn open-source BYOK Chrome extension that applies personal, reversible filters to the X timeline: you choose interests, weights and thresholds, Jev scores eligible post text remotely, and every display decision stays on your device.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
mithrilman/your-signalJavaScript
mithrilman/your-signalJavaScriptA bring-your-own-key Chrome extension that asks Jev one slop/not_slop Choice per X or LinkedIn post as it scrolls into view, then blurs and stamps anything at or above a configurable threshold. Scrolling never blocks, and each post is scored once from a local cache.
Jev Tools & CLIs#community#chrome-extension#moderation
Community
davertor/jev-slop-guardJavaScript
davertor/jev-slop-guardJavaScriptChrome extension that detects sponsored segments in Bilibili videos with TypeSafe Jev and skips them only when confidence passes conservative checks. A fork of ZeroTang05/VideoAdGuard, analyzing subtitles or Groq transcripts.
Jev Tools & CLIs#community#catalog-import#logicrw
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.