Hacker News new | ask | show | jobs
by tlb 876 days ago
The real prize is to deshittify the content on the web. Ignoring copyrights for a moment, it should be possible to rewrite pages in 1000s of sites down to just the information. StackOverflow could just have the Q&As, news sites could just have the news, Pinterest would just have the pictures. Remove all the signup nags and popups.

With good tools, one person could probably maintain the deshittifier for a few sites, at least until the sites started getting adversarial about it.

4 comments

ChatGPT is basically the deshittifier that rewrites thousands of sites to just the information…?

And it has a much better interface than original Google. Every query starts with the equivalent of “I’m feeling lucky”, but then you can ask further questions.

AI generated garbage is literally one of the biggest reasons search has turned to shit[0]. Even with SEO, Google was sometimes useful. Now most search results just return AI hallucinations.

[0]https://news.ycombinator.com/item?id=38952526

[1]https://aftermath.site/the-internet-is-full-of-ai-dogshit

That’s primarily Google’s problem though. ChatGPT was trained on web content before the AI deluge.
I would be a problem for any search engine focused on relevance, because this content is proliferating throughout the web, Google is poisoned because everything is poisoned.
Nah this tech will poison the web. People don't know what it is and take it too seriously.
SO is one of the examples where there isn't that much bloat, usually the longest answers are more detailed and let you learn, instead of just copy pasting (potentially) working code.

In the web Google is in part responsible for adding more content, they always suggested longer articles because it helped the algo get a better context. But for us humans it mostly doesn't make sense.

The same happens with books but I think we are at fault too, I would think twice about a book that is 80 pages, but the truth is that 80 pages could be a lot for most topics. I believe that the summarization capabilities of LLMs are gonna make a generation feel different about short content.

Nobody will create that. Too complicated to create. Too complicated to maintain.

I wish there was a browser add-on that could regex-replace html source of the page. Then I could write my own deshittfier list. Since I couldn't find any, i'm guessing it's because the plugin APIs won't allow it.

There was a MITM app that could do this, AdMuncher, but then the web switched to HTTPS and it didn't work anymore.

Reminds me of the word of the year https://en.wikipedia.org/wiki/Enshittification