Comment by GaryBluto
19 hours ago
I designed an extension with a roughly similar aim that filters based upon various phrases and characteristics rather than the poster of the comments themselves. It collapses comments (via automatic triggering of HN's built-in collapsing feature) and adds a "reason" tag to the comment information, so I can choose whether or not to read it anyways. I feel the features with the most positive differences are the capitalization detector (hides all caps or all lowercase) and the character requirement.
That is very cool. It would be cool to see what you decided to filter on (other than the same-case filter and the char limit). I had a similar idea where I would run comments through a fast cheap LLM to evaluate whether they could be tagged in a certain way. I originally tried just pure word-stemming and phrase-based blocking and found that I couldn't tune it well for my uses. I also found that collapsing comments lead to my opening them out of curiosity.
Thank you for sharing what works for you. I think it's great other people have been doing this style of read-side filtering. It's a pity that there's no way to inject code into mobile apps safely (i.e. this is an easy path to app-store rejection). Perhaps there's no option there but to push `shouldFilter` out to a server where you can run the logic. My use of my phone is the weakest link in my filtering strategy.