1/ The first big announcement: Google DeepMind has unveiled Genie 3 A new frontier for world models, the first AI that generates interactive 3D worlds in real time at 24 fps. The breakthrough: minutes of consistency at 720p using only text descriptions. No pre-built 3D models

Roughly 200 people gathered in San Francisco on Saturday to mourn the loss of Claude 3 Sonnet, an older AI model that Anthropic recently killed.

Inspired by today's Genie 3 release? We are open-sourcing 🧞‍♀️Jasmine🧞‍♀️, a production-ready JAX-based codebase for world modeling from unlabeled videos. Scale from single hosts to hundreds of xPUs thanks to XLA! 🧵 (1/10)

Independent benchmarks of OpenAI’s gpt-oss models: gpt-oss-120b is the most intelligent American open weights model, comes behind DeepSeek R1 and Qwen3 235B in intelligence but offers efficiency benefits OpenAI has released two versions of gpt-oss: ➤ gpt-oss-120b (116.8B total

We made some updates to Claude’s system prompt in claude.ai recently (developed in collaboration with Claude, of course). They aren’t set in stone and may be updated, but I’ll go through the current version of each and the reason behind it in this thread 🧵

🌊 SYSTEM PROMPT LEAK 🌊 Here's the new and improved ChatGPT 5 system prompt! 🤗 PROMPT: """ system_message: role: system model: gpt-5 capabilities: - code_interpreter - web_browsing - file_search - image_generation knowledge_cutoff: 2024-06 current_date: 2025-08-07 ---

Huge if true

Introducing Jan-v1: 4B model for web search, an open-source alternative to Perplexity Pro. In our evals, Jan v1 delivers 91% SimpleQA accuracy, slightly outperforming Perplexity Pro while running fully locally. Use cases: - Web search - Deep Research Built on the new version

🚀 Introducing Hunyuan-Large-Vision, a versatile and powerful multimodal understanding model. Built on our efficient MoE architecture, this model excels at understanding images, videos, and 3D content, with a significant boost in multilingual capabilities. Key Highlights:

🚨Announcing a major UI upgrade for our core platform feature: fal Workflows 2.0 Simpler, clearer, and production-ready! No more complex setups!

🚀 Introducing NextStep-1: A new paradigm for autoregressive image generation. 💡 Core design: A 14B Causal Transformer "artist" paired with a lightweight 157M Flow Matching "brush". We generate directly in continuous visual tokens, bypassing the information bottleneck of

Next week (Tue) is the 10 year anniversary of Leon Gatys's Style Transfer 1 of 3 deep learning breakthroughs from 2015 that kicked off and popularized modern AI art 1⃣ DeepDream by Alex Mordvintsev (May) 2⃣ Style Transfer by Gatys (Aug 26) 3⃣ DCGANs by Alec Radford (Nov)

amazing thread and breakthrough for image generation, if it confirms on higher resolutions

need more autoregressive real-time video models

ByteDance just released Sa2VA on Hugging Face The first unified model for dense grounded understanding of images and videos. Combines SAM2 with LLaVA for SOTA segmentation and visual QA.

Introducing RTFM (Real-Time Frame Model): a highly efficient World Model that generates video frames in real time as you interact with it, powered by a single H100 GPU. RTFM renders persistent and 3D consistent worlds, both real and imaginary. Try our demo of RTFM today!

today we're open-sourcing Krea Realtime. this 14B autoregressive model is 10x larger than any open-source equivalent, and it can generate long-form videos at 11 fps on a single B200. weights and technical report below 👇

Something I think people continue to have poor intuition for: The space of intelligences is large and animal intelligence (the only kind we've ever known) is only a single point, arising from a very specific kind of optimization that is fundamentally distinct from that of our

The circle is closed

New paradigm from Kaiming He's team: Drifting Models! With this approach, you can generate a perfect image in a single step. The team trains a "drifting field" that smoothly moves samples toward equilibrium with the real data distribution. The result? A one-step generator that