I truly have no words for what GPT Image 2.0 Is capable of... Like this was all from the model's memory. I assume it had to look up some information because theres no way it natively knew what Project Glasswing or Claude Opus 4.7 are
> be Anthropic
> train sonnet 5
> mfw it's better than opus and also faster
> send it to companies to announce release
> moneybags.gif
> rename to Opus 4.6
> slow it down
> undercut OAI 5.3 release
> wait
> release original Sonnet speed with an up-charge
> profit $$$
Our teams have been building with a 2.5x-faster version of Claude Opus 4.6.
We’re now making it available as an early experiment via Claude Code and our API.
Opus 4.6 is actually properly released on claude.com
I assume were going to have to wait at least a couple days for sonnet 5 then.
Were only a couple minutes away from a blog post by anthropic with benchmarks if the last stealth release is anything to go by.
I keep noticing Claude 4.5 opus mysteriously includes the word "mass" where it doesn't make sense. I kid you not it even randomly said "That's mass a mass mass!" And then it immediately apologized and said it seemed like it had a stroke. LMAO. Even after that, still a lot of random uses of the word where it doesn't make sense.
Now were talking, Qwen!
I love how the idea of seamless seperation for really any sort of image editing.
Photoshop is in trouble lol
youtu.be/OVhmiBrsziQ
This is such an interesting move. I feel like AI code IDEs were always built from the ground up for programmers (it is an IDE afterall)
But now Cursor is randomly making a Design -> Code interface that's better than 99% of existing products? Insanity
While actively working on a project in cursor; I could see in real time GPT-5's quality degrade (even in brand new chats) to the point where it would fail to edit a file 6 times in a row... I want to emphasize that this was happening first message in new chats! It's like they swapped the model under my nose or something, how does it consistently get worse? @cursor_ai@OpenAI
Open AI keeps shipping, but not having their new models integrate their existing features. Does anybody remember the Assistants API? 4o is the latest model that supports it. I understand that tool calling and system notes exist, but handling messages on the cloud was nice.