gpt-5 is the best coding model in the world and is now the default in Cursor. youtube.com/watch?v=PQUcIb…

Is Chain-of-Thought Reasoning of LLMs a Mirage? ... Our results reveal that CoT reasoning is a brittle mirage that vanishes when it is pushed beyond training distributions. This work offers a deeper understanding of why and when CoT reasoning fails, emphasizing the ongoing

Change my mind

Slackkkk

Acceleration>>speed

Genie 3 is amazing I think this is the most important release of the week. This should be the starting of Metaverse

Why is there so many mixed reactions on gpt-5. What's everyone's take ?

Users should be allowed to switched between models for reasoning and non-reasoning tasks

Gpt-5 is overhyped?

People use the model and say it's bad.. After few days People use the model and say it's SOTA

I let the llm pick the prompt

xAI Grok Updates (Last 24H): - Grok 4 boosts PDF handling for massive files! - iOS app v1.1.40 improves sound & Imagine. - Kids Mode hits Android soon. - longer vids in Imagine. - 44M images created, app #2 in Productivity! - Art Contest: Most-liked pics in X feed. - Tesla

I still love openai❤️ Sam Altman

Would you all think zucc poaching did affect openai gpt-5 release ?

Deepthink alpha

Lmaoo🤣

Fantastic paper from AI at Meta Reasoning LLMs hallucinate more on long answers, and the authors show why and fix it with a new reward that balances accuracy, detail, and relevance. Their online RL recipe cuts hallucinations by 23.1 points, raises factual detail by 23%, and

O4-mini is my favourite go-to model for all kind of tasks. With deciprication of gpt-4 models o4-mini type model became obsolete Sam Altman Greg Brockman

Large language models often sound sure even when they are wrong. This paper teaches a model to treat its own confidence as a training reward, which tightens calibration and improves reasoning without any human labels. Here is how it works. The model generates several chain of

arxiv.org/pdf/2508.05618