About defluffer

defluffer is a browser extension that rewrites long, AI-bloated LinkedIn posts into one honest line while you scroll. It is free, requires no account, and is available for Google Chrome and Mozilla Firefox on desktop.

Why it exists

LinkedIn rewards a particular shape of writing: eight paragraphs of preamble, gratitude, and reflection wrapped around a single fact. Generative AI made that shape cheap to produce at volume, and in 2026 a widely reported detector flagged roughly 41% of long-form LinkedIn posts as AI-generated. The result is a feed where reading a post costs far more attention than the information inside it is worth.

defluffer was built to invert that trade. Instead of asking people to write less, it changes what the reader receives: the one concrete thing a post actually says, stated plainly, in place of the wall of text.

What it does

• Replaces long feed posts with a single summarizing line as you scroll.
• Grades each post's fluff level and colors its badge accordingly, so pure engagement bait is visibly marked as such.
• Leaves the author, avatar, reactions, and comments exactly as they were. One tap restores the original post.
• Marks already-short posts with a green "fluff not found" badge. Those posts are never sent anywhere.
• Grades your own draft before you publish it, through a "check fluff" control in the LinkedIn post composer.

How it is built

A content script identifies long posts in the LinkedIn feed and sends the post text to a Cloudflare Worker. The Worker calls Google's Gemini 2.5 Flash-Lite with a tuned system prompt and returns a one-line summary along with a fluff grade of LOW, MEDIUM, HIGH, or PURE. Summary quality is maintained against a golden set of real feed posts that are rated by hand whenever the prompt changes, which is how the tone stays dry without becoming cruel.

The extension is open source under the MIT license. The full source, including the exact system prompt that ships in production, is available on GitHub.

Privacy

Nothing is stored. Post text is processed in transit to produce a summary and is then discarded. There is no database, no account, no login, and no usage analytics. Full detail is in the privacy policy.

Who made it

defluffer was built by Itay Zitvar, a product manager at NVIDIA and a Wharton MBA. Questions, bug reports, and feature requests are welcome at itayzit@gmail.com.

Home · About · Contact · Privacy · llms.txt · Source
Not affiliated with or endorsed by LinkedIn.