@theo Seems like ChatGPT desktop app is now Codex/Work first, with the “chat” part of it being an afterthought, existing as a little pop up, with all chat “Projects” gone, no more deep research, and you can’t select the model when you use chat.
Where did chat “Projects” go in the new unified ChatGPT app? Also, in chat mode, it seems you no longer can switch which model you want to use? And finally, what happened to Deep Research? @OpenAI ?
@tetsuoai This tweet is what happens when you don’t know the fundamental principles such as the difference between non-contiguous data structures such as Python’s lists and contiguous memory blocks that are C-arrays 🤷♂️
@martinmrmar In college I always thought one-column mostly text format was so much easier to focus on. The comic-book-like experience of modern calculus text books with lots of graphs, pictures, trivia, multiple columns, always felt harder to follow.
@theo Dude, this is a flash model! 😄 It sits between Opus 4.7-low and Opus 4.7-medium in terms of performance and price. Were you expecting it to have Opus 4.7-max performance with Kimi 2.5 pricing?
@svpino Wow, the fingerprints on that laptop are intense! Is it really that much of a fingerprint magnet? Or were you using a bag of chips as a mousepad the entire 1.5 hours at the airport? 🤔
@predict_addict You can’t really compare a quantized 35B-parameter distilled model to the likes of GPT-5, Gemini 3, or Sonnet 4. Also, that 50 TPS will drop sharply as soon as you have anything meaningful in the context. Open models are impressive, but they aren't quite there yet.
Sam Darnold’s comeback story over the past 2 years has been one of the most inspirational things to watch! Cool seeing him overcome!!! Congrats bro.
Hope you win it! 🏆
Go Hawks