Welcome to the "AI Daily" section! This is your guide to exploring the world of artificial intelligence every day. Every day, we present you with the latest content in the AI field, focusing on developers to help you understand technology trends and innovative AI product applications.

New AI products Click to learn more: https://app.aibase.com/zh

1. ChatGPT officially moves into Chrome browser: can browse web pages, manage tabs, and access your local files

OpenAI has upgraded the ChatGPT extension, allowing it to run directly in the Chrome browser sidebar without switching to a standalone app or website. The extension can browse web pages, read content, coordinate tasks across tabs, and access the local file system and installed plugins, marking the browser's evolution from an information display window to an AI execution workspace.

image.png

AiBase Highlights:

The ChatGPT extension now supports running in the Chrome browser sidebar, offering a native browser experience.

The extension can browse web pages, read content, coordinate tasks across tabs, and access the local file system.

OpenAI is shifting focus to embedding in mainstream browsers rather than maintaining standalone browser products to reach more users.

2. OpenAI releases GPT-5.6 model series: Sol, Terra, Luna versions launched, focusing on cybersecurity and cost-effectiveness

OpenAI released the GPT-5.6 model series, including Sol, Terra, and Luna versions, focusing on cybersecurity and cost-effectiveness. This series offers significant advantages in performance and cost efficiency, and introduced the enterprise office assistant tool ChatGPT Work, further expanding its influence in the B-end market.

image.png

AiBase Highlights:

🧠 The GPT-5.6 series model performs well in cybersecurity, covering key tasks such as threat modeling and code reviews.

📈 The Sol model shows significant efficiency improvements and cost advantages in AI coding tasks.

💼 OpenAI launched the ChatGPT Work tool, expanding its ecosystem in the B-end collaborative office field.

3. Anthropic launches AI mirror feature: let your AI assistant write you a usage health report

Anthropic's Reflect feature analyzes users' AI usage habits and provides personalized summaries to help improve their questioning skills. At the same time, the feature emphasizes privacy protection to ensure user data security.

image.png

AiBase Highlights:

🧠 The Reflect feature helps users review and examine their AI usage habits.

📊 Generates personalized usage summaries covering four dimensions: task delegation, goal description, etc.

🔒 Privacy protection is clearly defined, not reading incognito chat content.

4. Tencent to take over Manus as its largest shareholder, valuing 2 billion USD to develop the intelligent agent ecosystem

Tencent is considering acquiring Singapore-based AI startup Manus at a valuation of 2 billion USD, becoming its largest shareholder, which will help Tencent in its ecological layout in the AGI era.

image.png

AiBase Highlights:

🎮 Tencent is planning to take over Manus, valued at 2 billion USD, becoming its largest shareholder.

💼 Manus' founder has a deep connection with Tencent, who previously received investment from Tencent.

🚀 Manus plans to list in Hong Kong and may need to restructure to meet Chinese regulatory requirements.

5. ByteDance's Seedance 2.5 model will fully open API on July 16th

ByteDance's latest video generation model, Seedance 2.5, will fully open its API on July 16th, marking the transition of the model from gray-scale testing to full commercialization, which is expected to further lower the barrier for high-quality video generation. Seedance 2.5 has achieved technical breakthroughs in single-segment generation length, multi-material reference, and video editing, accelerating the engineering implementation of AI video tools in the film and television creativity and commercial advertising fields.

AiBase Highlights:

🎥 Seedance 2.5 supports 30-second single-segment native video output, enhancing the space for coherent expression of AI videos.

🔄 Enhanced multimodal integration capabilities, with reference material capacity increased up to 50 full-modal materials.

🔧 Introduces more controllable video generation and editing features, optimizing creators' precise micro-adjustment of image subjects and camera movements.

6. Pre-train from scratch, Ant Group's LingBot-VA2.0, a body-native world action model

The article details Ant Group's release of the embodied native world action model LingBot-VA2.0, which marks a critical shift in robot foundation models from digital world model building to physical world native design. Through self-regressive architecture pre-training from scratch, the model performs well in dynamic modeling, causal prediction, and real-time execution, solving the industry's common problem of low execution efficiency.

image.png

AiBase Highlights:

🧠 Introduces a semantic visual-action tokenizer, improving instruction following and action accuracy.

🔄 Uses a strict causal pre-training paradigm to ensure visual prediction and action generation follow a unidirectional time sequence.

⚡ Achieves real-time reasoning efficiency of 150Hz on a single card, solving the problem of low execution efficiency in embodied world models.

7. Completed a new round of financing of 16 billion HKD, founder Yan Junjie announced zero salary until achieving AGI

MiniMax completed a new round of financing of 16 billion HKD, and the founder Yan Junjie pledged to receive zero salary and invest shares to motivate the team and open-source community. Industry analysts believe that the financial reserves will help commercialization.

AiBase Highlights:

🚀 MiniMax completed a new round of financing of 16 billion HKD, used for AI infrastructure and model research and development.

💼 Founder Yan Junjie announced zero salary until achieving AGI, and will use shares to motivate the team and open-source community.

📈 Industry experts believe that the financial reserves enhance the ability to promote commercialization.

8. Samsung develops 4nm AI PC dedicated chip "GAIA" and has provided samples for testing to Lenovo and HP

Samsung Electronics is developing a generative AI acceleration chip called "GAIA," using a 4nm process technology, and has already provided prototype samples to PC manufacturers such as Lenovo and HP for testing. GAIA's core is a storage-centric architecture, aiming to deeply integrate computing units with memory while advancing the integration with next-generation DRAM technology to reduce data transmission latency and power consumption. Unlike companies like NVIDIA and Qualcomm, GAIA is positioned as an independent AI computing module that works in collaboration with existing PC platforms. Samsung hopes to achieve a breakthrough in the competitive AI PC market through this strategy.

AiBase Highlights:

🧠 Samsung develops 4nm AI PC dedicated chip "GAIA," aimed at improving AI computing efficiency.

🚀 GAIA uses a storage-centric architecture, integrating with next-generation DRAM technology to reduce latency and power consumption.

🤝 Samsung uses GAIA to connect logic chips and semiconductor storage businesses, attempting to break through in the AI PC market.