Inspiration
I genuinely felt depressed whenever I opened LinkedIn.
Scrolling the feed made me feel behind, jealous, and constantly comparing myself to other people’s highlight reels. What was supposed to be a platform for finding jobs had started to feel like just another social media app, full of content that had little to do with actually building a career.
This isn’t just a personal experience.
Research has shown that use of professional social networking sites like LinkedIn is associated with higher levels of depression and anxiety among young adults, even after accounting for age, income, and overall social media use.
Studies also show that frequent exposure to comparison-driven content can negatively affect mental health — especially for people early in their careers who are still figuring out their path.
Slop Block was built on a simple belief:
LinkedIn should support job searching and professional growth. Not make people feel worse for logging in.
What it does
Slop Block filters your LinkedIn feed to focus on high-signal, job-relevant content instead of engagement-driven noise.
It automatically detects and blocks posts such as:
- 🚫 Hiring announcements you don’t care about
- 🚫 Grindset / hustle culture posts
- 🚫 AI doomer content
- 🚫 Child prodigy posts
- 🚫 Engagement bait
- 🚫 Other low-signal content
Blocked posts are:
- Blurred (not deleted)
- Clearly labeled
- Manually revealable at any time
The result is a feed that feels:
Calmer. More intentional. Actually useful for job searching.
How to set it up
✅ YES — THIS IS AVAILABLE TO TRY
🚫 NO API KEYS REQUIRED
Head to:
https://imprayingonmylinkedinopsdownfall.tech/
…and follow the setup instructions.
How classification works
By default, Slop Block works out of the box using fast, local heuristics.
If you also enable Chrome’s local Google Gemini LanguageModel API, the extension can additionally label posts that are ambiguous and can’t be confidently classified by heuristics alone.
⚠️ Important note about first-time AI usage
The first time the local Gemini model runs, it may take a noticeable amount of time to respond.
This is expected.
- The model is initialized locally by Chrome
- On slower machines, this initial setup can take longer
After the first run, subsequent classifications are significantly faster
If the local Gemini API is not enabled
✅ Posts are still blocked correctly using heuristics
⚠️ Ambiguous posts will not receive an AI-generated description
✅ No functionality is broken
✅ No setup is required to use core features
All AI runs locally.
No API keys. No accounts. No external servers. No data leaves your browser.
How we built it
Slop Block is a Chrome extension built using Manifest V3.
It injects content scripts into LinkedIn and watches for new posts using a MutationObserver, since LinkedIn heavily virtualizes its feed.
Each post is identified using LinkedIn’s stable activity URNs, ensuring that:
- Posts are never processed twice
- Performance remains predictable
Classification pipeline
We use a heuristics-first approach:
- A local rules engine analyzes post text
- Text is normalized and matched against phrases and patterns
- A strict precedence order ensures deterministic classification
This handles most posts instantly without any AI calls.
For posts that can’t be confidently classified:
- We fall back to Chrome’s built-in Gemini LanguageModel API
- The model runs locally in a background service worker
- It is only invoked for ambiguous cases
No API keys. No servers. No external requests.
UI & performance
- GPU-accelerated blur effects
- Non-destructive overlays
- No layout breakage
- Smooth scrolling preserved
A popup UI lets users toggle which categories they want to see, with preferences synced via chrome.storage.
Challenges we ran into
The biggest challenge was getting the local Gemini model to run at a usable speed.
Early versions felt sluggish and disrupted scrolling. We had to:
- Aggressively minimize AI calls
- Cache model sessions
- Keep prompts extremely constrained
LinkedIn’s DOM also changes frequently and relies heavily on virtualization, which made post detection brittle.
Reliable filtering without breaking scrolling required careful batching and idle-time processing.
Accomplishments we’re proud of
- 🚀 Works almost entirely locally
- 🔐 No API keys, no backend, no data collection
- 🎯 High filtering accuracy via heuristics-first design
- 🧠 AI only used when truly necessary
We’re also proud of how native the extension feels. Fast, reversible, and fully under user control.
What we learned
- Heuristics are still incredibly powerful when designed well
- Most content can be classified deterministically. Faster and more predictably than AI
- Chrome extension architecture (MV3, service workers, DOM mutation handling) is non-trivial
- Local LLMs can be integrated safely without hurting UX
Most importantly:
Product ideas rooted in genuine emotional frustration resonate far more than purely technical experiments.
What’s next for Slop Block
- Fix remaining edge-case misclassifications
- Push local Gemini usage closer to near-instant responses
- Smarter caching and invocation rules
- User-defined custom filtering instructions
Long-term, we want Slop Block to adapt to your job search style, not force you into someone else’s feed.
Log in or sign up for Devpost to join the conversation.