Today:

• Perplexity Comet Shows Power of Autonomous Browsing

• China Unveils Global AI Action Plan

• Meta Appoints Shengjia Zhao Chief Scientist

• Anthropic Eyes $150B Valuation Boost

• Microsoft Tests New GPT‑5 Mode

Perplexity Comet: Autonomous AI Browsers Actually WORK!

The Comet browser, integrated with Perplexity AI, demonstrates advanced web automation, from unsubscribing emails and scheduling calendar events to ordering groceries on Instacart and summarizing podcasts. 

While it struggles with complex tasks like Photopea editing, its voice-command and multi-agent capabilities hint at a future where AI handles routine online tasks seamlessly. Privacy concerns remain, but its Chrome-like interface and task personalization show strong potential for everyday productivity automation.

WATCH THE VIDEO ON YOUTUBE

China Unveils Global AI Action Plan

China has launched a global artificial intelligence (AI) action plan, emphasizing worldwide cooperation and proposing a global AI organization. This move follows a U.S. AI plan announced by President Trump, which focuses on reducing bias and expanding American technology. China aims to integrate AI across industries and support developing nations, while both countries build opposing alliances. The rivalry highlights efforts to control AI development, chip access, and international technology leadership.

Why It Matters

1. Global AI Alliances: China’s proposal for a global AI organization challenges the U.S.-led approach, potentially splitting global standards and partnerships.
   

2. Tech Power Competition: The U.S.-China race for AI dominance will shape the future of chips, data control, and advanced model development.
   

3. Access for Developing Nations: China’s support for the Global South could broaden AI adoption and shift economic and technological influence.

Meta Appoints Shengjia Zhao Chief Scientist

Meta appointed former OpenAI researcher Shengjia Zhao as Chief Scientist of its new AI unit, Meta Superintelligence Labs (MSL). Zhao, a key contributor to ChatGPT, GPT-4, and OpenAI’s reasoning model o1, will shape MSL’s research strategy under Alexandr Wang. Meta is aggressively recruiting top AI talent, boosting cloud infrastructure, and preparing its 1-gigawatt Prometheus cluster to train frontier models, aiming to rival OpenAI and Google in next-generation AI.

Why It Matters

1. Talent Shift: Meta’s recruitment of top OpenAI and DeepMind researchers signals intensifying competition for AI expertise.
   

2. Reasoning Models Focus: Zhao’s leadership suggests Meta is prioritizing advanced reasoning AI, challenging OpenAI’s early lead with o1.
   

3. Massive Compute Investment: The upcoming Prometheus cluster reflects the escalating scale and power race needed to train frontier models.

Anthropic Eyes $150B Valuation Boost

Anthropic, a leading AI start-up and OpenAI rival, is in talks with Middle Eastern investors to boost its valuation to over $150 billion. Historically wary of Gulf sovereign funds due to ethical concerns, Anthropic stresses responsible AI and alignment with human values. The company’s funding strategy aims to accelerate AI development while sparking debates on governance, societal impact, and the risks of rapid advancements in frontier AI technologies.

Why It Matters

1. Massive Funding Impact: A $150 billion valuation could fast-track Anthropic’s AI research and intensify competition with OpenAI.
   

2. Ethical Investment Debate: Anthropic’s cautious stance on Gulf funding highlights growing concerns over AI governance and investor influence.
   

3. AI Governance & Oversight: The deal raises questions about balancing rapid AI advancements with ethical frameworks and societal safety.

🧠RESEARCH

Group Sequence Policy Optimization

Group Sequence Policy Optimization (GSPO) is a new training method for large language models. Instead of focusing on individual words, it looks at whole sentences to adjust rewards and learning steps. GSPO is faster, more stable, and improves performance compared to older methods, boosting Qwen3 model results.

nablaNABLA: Neighborhood Adaptive Block-Level Attention

NABLA (Neighborhood Adaptive Block-Level Attention) speeds up video generation by reducing the heavy processing load of transformers. It focuses attention on key blocks of video data instead of every detail, using an adaptive threshold to save time. NABLA is up to 2.7× faster with minimal quality loss.

MUR: Momentum Uncertainty guided Reasoning for Large Language Models

Momentum Uncertainty-guided Reasoning (MUR) makes large language models think more efficiently. It tracks uncertainty at each step and spends more effort only when needed, cutting wasteful calculations. With a simple control setting, MUR improves accuracy by up to 3.37% while reducing computation by over half on key reasoning benchmarks.

🛠️TOP TOOLS

ArtGuru Face Swap - AI tool designed to make face swapping in photos an effortless and enjoyable process.

Nokemon - A tool that leverages advanced ML technology to create unique and customizable Pokémon designs, often referred to as “Fakémon.”

NeuralBlender - Image-generation website that leverages the power of AI to create stunning images from textual descriptions.

CrushOnAI - AI chat platform that specializes in providing unrestricted, NSFW (Not Safe For Work) conversations.

Charley AI - AI-powered content generation platform designed to assist with academic and professional writing.

📲SOCIAL MEDIA

🗞️MORE NEWS

• Microsoft may soon add GPT-5 to Copilot through a new “Smart Mode” that can think fast or in detail. References to this feature were found in Copilot’s code. OpenAI plans to launch GPT-5 in early August.

• Huawei unveiled its CloudMatrix 384 AI system at the Shanghai WAIC, aiming to rival Nvidia’s top product. Using 384 of its 910C chips, Huawei claims superior performance through advanced system design and high-speed “supernode” connections despite weaker individual chips.

• Alibaba unveiled its first AI-powered Quark glasses at Shanghai’s WAIC, marking its entry into China’s competitive smart wearable market. The glasses integrate Alibaba’s AI ecosystem, offer payment and navigation features, and use a Snapdragon AR1 chip for improved performance and efficiency.

• Meta appointed Shengjia Zhao, ChatGPT’s co-creator, as chief scientist of its Superintelligence Labs, prompting questions about Yann LeCun’s role. LeCun clarified he remains Meta’s chief AI scientist at FAIR, focusing on long-term AI research, while Zhao leads superintelligence efforts.

• University of Tokyo researchers used an AI system called VBayesMM to map connections between gut bacteria and the chemicals they produce. This breakthrough could lead to personalized treatments, targeting bacteria to improve digestion, immunity, and mental health.

• Sam Altman warned that ChatGPT lacks legal confidentiality, unlike conversations with doctors or lawyers. He said OpenAI could be required to share private chats in lawsuits, stressing the need for stronger privacy rules before widespread AI therapy use.