AI Daily.

The Rolling Forcing method, jointly developed by Polytechnic University and Tencent ARC Lab, solves the problem that it is difficult to balance the quality, consistency and real-time of AI long video generation. The method uses rolling window joint noise reduction, Attention Sink mechanism and efficient training algorithms to achieve 16 fps minute-level high-quality video stream generation on a single GPU, effectively suppressing error accumulation and screen its support for interactive dynamic guided content creation, and the related code and model have been open-sourced.

Cursor 2.0 is officially released with the launch of Composer, a self-developed large model that generates code at a speed of up to 250 tokens/second, two times faster than GPT-5 and Claude Sonnet 4.5. The model is specially trained for real development scenarios and can autonomously complete the whole process of coding, testing and bug fixing, and is currently only integrated into the Cursor editor for internal use.

Cursor 2.0 is officially released with the launch of Composer, a self-developed large model that generates code at a speed of up to 250 tokens/second, two times faster than GPT-5 and Claude Sonnet 4.5. The model is specially trained for real development scenarios and can autonomously complete the whole process of coding, testing and bug fixing, and is currently only integrated into the Cursor editor for internal use.

FlowithOS is the world's first native operating system designed for AI intelligences, capable of performing complex tasks rather than being limited to chatting. Based on the Chromium browser, the system supports cross-platform multitasking in parallel with unlimited contextual memory and Skills library, and can automate real-world tasks such as Taobao shopping, microblogging and data collection. Currently in early beta, it supports Windows and macOS, and requires an invitation code to experience.

FlowithOS is the world's first native operating system designed for AI intelligences, capable of performing complex tasks rather than being limited to chatting. Based on the Chromium browser, the system supports cross-platform multitasking in parallel with unlimited contextual memory and Skills library, and can automate real-world tasks such as Taobao shopping, microblogging and data collection. Currently in early beta, it supports Windows and macOS, and requires an invitation code to experience.

Musk unveiled xAI's latest AI model, Grok 4, on July 10th, trained with 200,000 H100/A100 GPUs and breaking 50% accuracy in HLE tests. The model excels in several benchmark tests and is particularly well suited for complex reasoning tasks. The commercialized version of SuperGrok is priced at $30 to $300/month and is aimed at high-end professional users.Grok 4 will be integrated into eco-products such as Tesla & Optimus Robotics.

OmniGen2 is a multimodal generative model based on the Qwen-VL-2.5 architecture with 7 billion parameters, of which 3 billion are used for text processing and 4 billion for image diffusion generation. Its core capabilities include intelligent text-to-image, context-aware editing and multimodal understanding. The added self-reflection mechanism can autonomously optimize the output quality. With ComfyUI's node-based integration, users can operate it intuitively and lower the threshold of use. Professional-grade image generation and editing effects have been demonstrated in multiple scenarios.

GPT-5 will integrate several AI tools such as Codex and Operator to realize the integration of programming, research, operation and memory functions. It is fully multimodal and can handle voice, image, code and video inputs, and can intelligently switch between inference and dialog modes. According to tests, its programming efficiency can be increased by 3 times, positioning it as a key breakthrough in the third phase of AGI development. It is expected to be released within this year, triggering industry concerns and security discussions.