I built it because I was tired of social media websites pushing algorithmic content at me, while preventing content I actually subscribed to from being shown.

e.g. My wife has an art channel on YouTube with over 200K followers, yet I never see her videos on my feed despite being subscribed.

So, I created a browser extension that is an observability tool for algorithmic manipulation. It measures and gives visibility into how much of the feed is intentional, i.e., actively chosen, and how much is pushed onto you algorithmically.

==== Repo + installation ====

AttentionGuard: https://github.com/agentkites/attentionguard-extension

==== What AttentionGuard is ====

- Free and open-source browser extension (MIT) for Chrome and Firefox

- Analyzes social media and e-commerce feeds in real time

- Shows WHY each piece of content is there

- Everything runs locally in your browser, and no data or telemetry is collected

- Supported sites: Reddit, Twitter / X, FB, Insta, LinkedIn, YouTube, and Amazon

==== How AttentionGuard works ====

As you browse supported sites normally, it classifies feed items into:

- Ads: sponsored or promoted content

- Algorithmic: recommendations selected by the platform

- Social: shown because someone engaged with it

- Organic: content from accounts you actually follow

Click the extension icon to see a live breakdown of your session.

==== Limitations ====

- All detection is based on visible DOM labels, attributes and page structures - not internal platform API’s or algorithms.

- Platforms frequently change UI structures, so detection may break until updated

- Currently only works on homepage feeds of the supported sites

- We're working on getting it listed in extensions marketplace in Chrome/Firefox.

==== What we need ====

- bugs and breakage reports

- PRs for new platforms

- critiques of classification heuristics

Do share your questions and feedback!