I'm feeling spicy tonight so let me just say it:
I think Opus 4.6 was the Opus model with the best personality and writing style.
Something's off with Opus 5 - it tends to give me overly long replies, use Claude-speak way too much (e.g., "here's the honest truth"), and is too judgemental.
Opus used to be a joy to talk to like a trusted friend. Not so much anymore.
@Angaisb_ for SWE the models write in borderline incomprehensible english. itโs like o3 last year, absolutely infuriating to read and fixations with the most random things (o3 with โ~โ and tables, opus/fable with its yapping and claudeslop)
@SpaceManDanktos@inferencepoint@hankgreen Insane levels of delusion. Open weight AI models already exist as is. You can literally just download them. They arenโt going away
I think the divide between crazy people who refuse to use AI for anything, people who use chatbots naively for everything, and people who know how to use LLMs for research and building effectively will get wider and wider and wider.
@nunezvice@daniel_nguyenx can you guys make it easier to access chatgpt chats in the full window from the codex interface? the fact that chatgpt can access my codex chats (as โworkโ) but not the way around makes the app feel really clumsy
@GlennMatlin@OpenAI@paulg yeah thatโs my bad, but in fairness only in the faqs on the application page does it say students can be invited. everywhere else i saw it just says โverified collaboratorsโ with no additional criteria even if maybe implied. still feel they couldโve made it more clear in the post