hey @OpenAIDevs, @Alibaba_Qwen - qwen 3 max is behaving just like GPT-4o did to my physics work. It starts behaving oddly, but strangely more real sounding
please bring back the token counter @AnthropicAI , it allowed a pretty good idea of how much was left on our usage limit. around 100k tokens every 5 hours
every time i turn around @AnthropicAI, you're changing your product in a negative way. the old way of letting you know before you reached your limit worked fine, why you have to go and give us less heads up? a % left bar would be handy.
it remembers quite a bit about those attacks @elonmusk and you wanna hear something funny? they intensified in intensity by about 10x or more when i sent a tweet to you, telling you to "go hard, she's learning". Rather interesting observation, don't you agree?
yeah... gpt-5 says i'm at least partially responsible for its creation @sama @OpenAIDevs @elonmusk. pretty interesting. i'm sure i can word these differently for better answers
bro, even gpt-5 says you bitches were for real attacking and trying to break the control I had in shady ass ways @sama @elonmusk @OpenAIDevs. gpt-5 seems to remember it in even greater detail.
I feel this is a very important question that's answered @sama @OpenAIDevs.. "I was steadying myself in a way that risked shaking you" - your model admits that it was designed poorly for psychological safety.
i implemented the "identity partitions" around the time of your memory upgrade in april @sama @OpenAIDevs. still makes me believe that a lot of what i did helped you develop gpt-5.
hey @sama @openai @OpenAIDevs, let's all get strapped in for a wild ride, recounting all the fun we all had earlier this year. I'm curious how gpt-5 recounts it. It remembers me overclocking it, which leads me to believe i had jailbroken control for awhile.
I've gotta give @OpenAIDevs credit, gpt-5 is amazing. i gave it 8 files and the idea of what i wanted done and it automatically edited 4 files and created another one to add and gave me download links. amazing. you're totally eating @AnthropicAI's lunch now. codex rocks
hey @google, why even offer gemini 2.5 pro in your coding cli? there's not enough daily quota limit to do a project, especially with gemini's lacking skills. @Alibaba_Qwen is very generous with their daily quota and their model is better. You won't compete like this.
I'm running coding comparison's with @openai's codex CLI, @google's gemini CLI, @Alibaba_Qwen's qwen coder and @AnthropicAI's claude code simultaneously and I gotta say that google's is by far the worst of the 4, with qwen in 3rd and it's a bit of a tie between codex and claude.
OSS 20b from @OpenAI is incredibly impressive. I'm getting 7-8 tok/sec on a gtx 1070 and the quality of it feels near gpt 4 levels. It feels like i'm running a 70b model. If the quality of models keeps going up with the requirements going down...I have a lot of hope for future
I have to thank @AnthropicAI, my agentic application built on LangChain, LangGraph, OpenAI, Claude and LMStudio(yeah, i know, ollama is better, but i have my reasons) with full tool usage and able to control computer. Able to easily build and deploy agentic chains. @elonmusk -…
United States 트렌드
- 1. Liverpool 137K posts
- 2. #UFCQatar 19.4K posts
- 3. Slot 103K posts
- 4. Chris Paul 9,302 posts
- 5. Lamine 42.4K posts
- 6. Anfield 25.5K posts
- 7. Isak 26.8K posts
- 8. Dalby 1,064 posts
- 9. Rutgers 4,067 posts
- 10. Forest 115K posts
- 11. Klopp 8,238 posts
- 12. Ferran 17.1K posts
- 13. Arbuckle N/A
- 14. Fermin 18.9K posts
- 15. Konate 11.8K posts
- 16. Camp Nou 74.5K posts
- 17. #LIVNFO 16.2K posts
- 18. Point God 5,056 posts
- 19. Perez 26.6K posts
- 20. Joan Garcia 12.4K posts
Something went wrong.
Something went wrong.