[tolerance]

For things like coding LLMs are useful and DEVONThink's recent AI integrations allow me to use local models as something like an encyclopedia or thesaurus to summarize unfamiliar blocks of text. At best I use it like scratch paper.

I formed the habit of exporting entire chats to Markdown and found them useless. Whatever I found of useful from a given response either sparked a superseding thought of my own or was just a reiteration of my own intuitive thoughts.

I've moved from ChatGPT to Claude. The results are practically the same as far as I can tell (although my gut tells me I get better code from Claude) but the I think Anthropic have a better feel for response readability. Sometimes processing a ChatGPT response is like reading a white paper.

Other than that, LLMs get predictable to me after a while and I get why people suspect that they're starting to plateau.

[tombarys]

You are right. It plateaued and even degraded in some way. Or we just got more sensitive to its bullshiting?

[tolerance]

A little bit of both, I think. And I suspect that we aren't going to see another noticeable leap forward until specialized models become commonplace and/or people figure out exactly how LLMs productively fit their interests.