Slop Shame

Your LinkedIn feed, graded for AI slop. Then destroyed. Then shamed.

Slop Shame is a browser extension that gives every post on your LinkedIn home feed a Slop Index from 0 to 100: how much it reads like a language model wrote it after one prompt. If it's bad, you can blow it up. If it's really bad, you can roast its author.

It runs entirely on your machine. No servers, no AI, no accounts. Just regexes with opinions.

What it does

The Slop Index

Every post gets a score and a tier badge. Hover it to see exactly which tells gave the post away ("It's not X, it's Y", "Agree?", seven em dashes, "delve").

	Tier	Score	Meaning
	No Slop Found	0–19	No AI tells found. That doesn’t prove a human wrote it.
	Faint Whiff	20–39	Something smells a little… optimized.
	Certified Slop	40–59	Grade A, farm-raised LinkedIn slop.
	Slop Swamp	60–79	Wading waist-deep in thought leadership.
	SLOPOCALYPSE	80–100	Pure, uncut, machine-grade slop. Evacuate.

Destroy

Press it and the post erupts off your screen. Gone. It only disappears for you; the author's 14 reposts are unaffected.

Shame

Generates a roast card for the post's author: a wanted poster, a citation, a certificate of achievement in slop. It quotes their actual slop back at them, so every card is personal. Save it as a PNG and do with it what your conscience allows.

Hall of Slop

A local leaderboard of the people who post the most slop in your feed. Post it once, you're a statistic. Post it every morning at 8:03, you climb the ranks. Posts are counted once, however many times you scroll past them.

Install (from source)

Not on the Chrome Web Store yet. Build it yourself; it takes a minute.

git clone https://github.com/prashantbhudwal/slop-shame.git
cd slop-shame
pnpm install
pnpm build

Then load it in any Chromium browser (Chrome, Edge, Brave, Arc, …):

1. Open chrome://extensions (Edge: edge://extensions).
2. Turn on Developer mode.
3. Click Load unpacked and pick the .output/chrome-mv3 folder.
4. Open your feed and brace yourself.

Requires Node.js 22+ and pnpm.

Privacy

Everything stays on your machine.

• No data collection, analytics or telemetry.
• The extension makes no network requests of its own. Scoring is local pattern matching, not an AI service.
• Scores and the Hall of Slop live in the extension's own local storage. Delete them with Clear all in the Hall of Slop, or by uninstalling.
• It only runs on linkedin.com.

The full policy is in PRIVACY.md.

How scoring works

The Slop Index is a heuristic, not an AI detector. It looks for the tells AI writing leaves behind: overused model vocabulary, template phrases, "it's not X, it's Y" pivots, engagement-bait closers, emoji bullet lists, suspiciously uniform sentence rhythm, leaked chatbot artifacts like [Your Name] or utm_source=chatgpt, and invisible characters from "humanizer" tools.

Each tell has a weight (faint, clear, or smoking gun). The engine adds them up, adjusts for how dense they are in the post, and squashes the result into 0–100. One "leverage" in a long post is forgiven; a 60-word post with nine tells is not. Rules are organised in pluggable packs, compiled once, and the whole thing scores a 300-word post in under a millisecond.

It will be wrong sometimes. It's tuned hard against false positives: being accused of slop when you typed it yourself is its own kind of tragedy. When it's wrong, please report the score.

Many AI-tell patterns are adapted from avoid-ai-writing (MIT, © Conor Bronsdon) and humanizer (MIT, © Siqi Chen). See THIRD_PARTY_NOTICES.md.

Contributing

Found a phrase only a model would write? Adding it is a few lines of data and a test. Start with CONTRIBUTING.md and the rule pack guide.

License

MIT. Slop responsibly.