3 ms·

I just created a regex-based content blocker that literally blacks out text content on any webpage. Chrome extension. 2 modes - block only matches, or, block s

I just created a regex-based content blocker that literally blacks out text content on any webpage. Chrome extension. 2 modes - block only matches, or, block surrounding text (too). Create as many rules as you like, it will block anything. I entered it into Google Built-in AI hackathon the other day. Uses local LLM to scan blocked content to determine if regex was too trigger happy, then lets user unblock.

ZYbCRq22HbJ2y7 2y ago
What local LLM? What kind of prompt? That seems like a lot for what sounds like sentiment analysis?
- purple-leafy 2y ago
  Google Gemini Nano Not really sentiment analysis purely, though thats something I want to add. When I'm on hackernews, I'm tired of coming across comments about: - Donald Trump, Xi, Zelensky, Putin, Musk (Other people) - China / Russia / Ukraine / USA etc - Alcohol/drugs/gambling etc So I basically block all that content
  - esperent 2y ago
    Did it block your own comment after you made it? More seriously, any plans to share it? I've been waiting for someone to build a tool like this, although I expected to use sentiment analysis directly rather than regex. Personally I'd love to block nitpicking comments here on HN, not sure if that could be done with regex though. Edit: just saw you shared it in another reply.
    - purple-leafy 2y ago
      Yep shared in other comment, keep an eye on repo - will have updates in new year post hackathon judging
theaussiestew 2y ago
Could you share? I'm looking to use something like this.
- purple-leafy 2y ago
  Here ya go, build from source :) [0] Needs some tidying up, and you need to figure out how to run the "AI" by following this article [1] [0] - https://github.com/con-dog/context-aware-pattern-blocker https://github.com/con-dog/context-aware-pattern-blocker [1] - https://developer.chrome.com/docs/extensions/ai/prompt-api https://developer.chrome.com/docs/extensions/ai/prompt-api

ZYbCRq22HbJ2y7 2y ago

What local LLM? What kind of prompt? That seems like a lot for what sounds like sentiment analysis?

purple-leafy 2y ago
Google Gemini Nano Not really sentiment analysis purely, though thats something I want to add. When I'm on hackernews, I'm tired of coming across comments about: - Donald Trump, Xi, Zelensky, Putin, Musk (Other people) - China / Russia / Ukraine / USA etc - Alcohol/drugs/gambling etc So I basically block all that content
- esperent 2y ago
  Did it block your own comment after you made it? More seriously, any plans to share it? I've been waiting for someone to build a tool like this, although I expected to use sentiment analysis directly rather than regex. Personally I'd love to block nitpicking comments here on HN, not sure if that could be done with regex though. Edit: just saw you shared it in another reply.
  - purple-leafy 2y ago
    Yep shared in other comment, keep an eye on repo - will have updates in new year post hackathon judging

theaussiestew 2y ago

Could you share? I'm looking to use something like this.

purple-leafy 2y ago
Here ya go, build from source :) [0] Needs some tidying up, and you need to figure out how to run the "AI" by following this article [1] [0] - https://github.com/con-dog/context-aware-pattern-blocker https://github.com/con-dog/context-aware-pattern-blocker [1] - https://developer.chrome.com/docs/extensions/ai/prompt-api https://developer.chrome.com/docs/extensions/ai/prompt-api