User repository on GitHub
A Chrome extension that brings TypeSafe's Jev to X.com to analyze posts as you browse
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Demos & Experiments
Barrunto
A Chrome extension that asks Jev closed yes/no questions about the pages you browse and hangs a label on each item from a rule pack, with per-site sensitivity, read-ahead, optional fading or hiding, and a tuning mode that shows how every judgment was reached.
- Stars
- 2
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- elpumberto
- Last verified
- Oct 2026
About this project
A Chrome extension that reads what you browse, asks Jev about each thing it reads and hangs a label on it when it has something clear to say. When it is not sure, it stays quiet. Barrunto is Spanish for a hunch, and that is all a label is: a model's opinion, not a verdict on what was written or on whoever wrote it. What it reads and what it says depends on the site. Everything Barrunto knows about one thing to do on one site is a rule pack, and it ships with three:
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- One Jev call per item: nine yes/no trait questions for a post or comment, seven for a profile plus the kind of each of its posts.
- Code combines the answers through weighted recipes, a lukewarm answer counts for nothing, and the outcome is a strength from 0 to 1.
- Three packs ship: X posts (Bait, Flame, Signal), Hacker News comments (Insight, Snark, Tangent) and Jetective Jev profiles.
- Sensitivity runs from Low to Ultra and reuses stored answers, so moving the dial asks Jev nothing new.
Watch out
The README describes this as an experiment that is not in the Chrome Web Store and gets things wrong, and a label is explicitly a model's hunch rather than a verdict on what was written or on whoever wrote it. Installation is manual: download the zip from the latest release and load it unpacked with developer mode, or build on Node 22 or later and load the output folder, whose leading dot means the file picker may hide it unless you press Ctrl+H on Linux or Cmd+Shift+. on a Mac. Enabling a pack makes Chrome grant host permission for that one site and nothing else. Cost is easy to underestimate: read-ahead analyzes the next few items as well as what is on screen and every one of those calls is billed whether or not you read the result, and at read-ahead 0 an item is only analyzed once it has lingered on screen rather than flying past. Noise labels can be faded or hidden, and hidden items fold behind a line that says what and why with a Show control, because a hunch can be wrong. Jetective Jev judges what an account does and never who is behind it, skips protected accounts and ones that are not there, and only sees the posts X keeps near the viewport, re-asking as more appear up to thirty. With X posts on it also reads your own drafts, and each check is another billed call, with a reply judged without the post it answers. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
Labeling text, and the README is firm about the difference. A label is a model's opinion, not a verdict on what was written or on who wrote it, and everything uses looks like rather than is. The profile pack judges what an account does, never who is behind it, and no profile label counts as noise, so nobody is faded or hidden for what their account looks like.
More than the visible count suggests. Along with what is on screen it analyzes the next few items so labels are ready when you get there, and every one of those is paid for whether you read it or not. At read-ahead 0 an item is only analyzed once it has stayed on screen for a moment. Every draft check on X is also a call, billed like any other.
No. The answers are already stored, so moving sensitivity asks Jev nothing. The tradeoff is accuracy: on Low only clear cases get a label, each step up labels more and gets more wrong, and Ultra labels at the faintest hunch.
No, it is not published there. Download barrunto-chrome.zip from the latest release, unzip it, turn on developer mode in chrome://extensions and load the folder unpacked, or build it yourself on Node 22 or later and load the output folder the same way. That folder starts with a dot, so the picker may hide it until you press Ctrl+H on Linux or Cmd+Shift+. on a Mac.
There is no server. You paste your own TypeSafe key into the popup and the extension talks straight to the Jev API. Turning a pack on makes Chrome ask for host permission for that one site and nothing else, and turning it off hands the permission back. Ads, posts from protected accounts and comments that were flagged are left alone entirely.
A single call per item carries the item and the pack's yes/no trait questions, and each answer comes back as the probability of a yes. Code then combines them through the pack's recipe: weighted traits plus signals worked out from the page or by counting Jev's answers. A lukewarm answer is Jev not knowing and weighs nothing, the result is a strength from 0 to 1, and a label goes up only when that clears the threshold. Where labels disagree, the strictest wins.
Only what anyone can see on the page: since when the account is on file, whom it follows, how much it posts, and what kind each post is, from the author's own things to adverts or stock material. It looks again each time roughly half a dozen more posts appear, up to thirty. Protected accounts and ones that are not there are skipped, every line of the case file is a sentence picked by code, and it never draws a conclusion about who is behind an account.
More Jev resources like this
98GitHub stars
Self-hosted Docker service and web UI that packages Laya structured-decision inference behind a compatible System One endpoint.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
1Panel-dev/laya-serverPython
1Panel-dev/laya-serverPython23GitHub stars
Browser drone-fleet lab where TypeSafe Jev steers each drone as fast reflexes and an optional planner offers one-use strategy advice when confidence drops, without pausing control.
Jev Demos & Experiments#community#jevbest#catalog-import
23GitHub stars
Chrome extension that sends each X post to Jev with six typed questions in one call near the viewport, then renders a behavior chip under the post plus a session cost and latency panel.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
oso95/x-scannerTypeScript
oso95/x-scannerTypeScript14GitHub stars
Manifest V3 Chrome extension in TypeScript that sends each LinkedIn post to Jev and collapses bait, self-promo, off-topic, and sponsored posts into a one-line bar with reasons and scores.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
stefw/lkcleanTypeScript
stefw/lkcleanTypeScript7GitHub stars
Demo that turns free-text UI descriptions into real shadcn components using only Jev typed Choice, Noul, and Score answers, with no generated code on the decision path.
Jev Demos & Experiments#community#jevbest#catalog-import
8GitHub stars
Claude-driven browser automation where plain-English chat drives Brave or Chrome, Jev makes on-page decisions, and a guard freezes irreversible clicks.
Jev Demos & Experiments#community#jevbest#catalog-import
Community
ranjan2829/AskJevTypeScript
ranjan2829/AskJevTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.
Jev Demos & Experiments#gaming#agents#control
Read article
From the Jev community
Posts from builders shipping with Jev right now.