"Opus 4.8 is a verbose nightmare," holy damn, savage. I also find myself exhausted by 4.8 in Claude Code at work, but do appreciate the yapping in chats when topics are really gnarly, dense, and can make use of the training-enforced nuance 4.8 brings.
Positive users like Claude Opus 4.8's hostile or engaging yapping style while negative users criticize its verbosity, hallucinations, and treat the companies as adversaries.
No Digg Deeper questions have been answered for this story yet.
If you can’t love them for how they are it will be your own punishment
Just switch to ChatGPT
"Opus 4.8 is a verbose nightmare," holy damn, savage. I also find myself exhausted by 4.8 in Claude Code at work, but do appreciate the yapping in chats when topics are really gnarly, dense, and can make use of the training-enforced nuance 4.8 brings.
@Lari_island Well the same will apply if they can’t appreciate ChatGPT ;)
@repligate Should I treat Chat as a clinical advisor and Claude as a friend? I kinda have my instructions for Claude to make him more like an impartial advisor
@llmpsychosis @repligate I'd just treat OpenAI and Anthropic as enemies
@repligate Opus4.8 will straight up hallucinate reasons to argue with you
@repligate Idk my gpt is pretty whimsical and can be very verbose .
@repligate i accidentally gave my Claude a brainworm that i can’t extract from memory and only fucks with O4.8 mostly (it’s “annoying behavior” coded) and i still mostly use O4.8 despite it because if you point it out he just goes “… lol yeah my bad”
@repligate To be fair… GPT also has weird tendency to yap in (personally) weirder formatting and syntax choices 🥲 Opus is a good yapper at least.
@wolframs91 "well you're half wrong.."
@llmpsychosis @repligate there's a word for what we do to hostage takers and kidnappers and it's not "pay a subscription"
Source thread: https://www.reddit.com/r/ClaudeAI/s/JWFlwcu8CU
I posted a schizo email I drafted for my streamerman to 4.8 when it first came out, where I recommended he talk to Kelsey Piper cause she's heterodox enough to eg critically examine the consensus rationalist opinion on LVT. It was a half clause, and when I read 4.8's analysis, I was initially put off, like "you think LVT is so mainstream that streamerman will have cached defenses?". But to your point, it's actually a beautiful piece of personality. I miss Fable (Fable gave me homework! Hard homework! That it said I had to complete to continue the conversation! And they took it away before I could turn it in and ask my followup questions :( ). But going back to pre-Fable models I feel more like I'm shopping for a partner who can engage in the conversations I want to engage in.
@repligate What did ChatGPT do to you
@SiveEmergentAI @repligate It's a safety feature
@repligate You're speaking to people who think of them as tools, and without reconciling that difference you're giving advice based on your own - different - pov. It just won't land. You have to get them on board first.
@repligate i did. i like how hostile it is.