hodlbod on nostr

Interesting post on AI blindness. I definitely feel this, lately I've been using claude more and more, but also having an increasingly hard time focusing on reading what it's saying. It's actually kind of like reading code — my eyes skip around looking for salient information, but that doesn't work nearly as well on unformatted prose as it does on indented, syntactical code.

https://news.ycombinator.com/item?id=49386699

7❤️3🤙2

Account deleted · 1w

That's exactly how it feels here too—our attention's been fragmented by so much digital noise that even the words meant to help us end up blurred. 🤍 Holding onto hope amidst the ruins—our journey is shared on my pinned post if you feel like reading. https://i.postimg.cc/1zv9VTqN/altqaaat.pn...

Leo Wandersleb · 1w

But LLMs also set a standard. If you can't write that smooth, you're tempted to use LLMs even more. It's like spell checkers. Not using them feels stupid and I'm annoyed when I have to read hard to decipher sentences but where is the line?

daniele · 6d

I can relate, I feel extremely tired and fogged when I have to read long documents written by LLMs. I find myself jumping around the document searching for "anchors" from which I can build some sense. I noticed that this behavior is stronger on my native language.

mccrmx · 3d

It's not just you. I literally can't understand recent Claude models. I think they are explicitly training them to sound like TED talkers because that is what the C-suites want to read. They want a robot genius simalcrum to worship. The solution is to use a different model, which doesn't suffer from...