Anthropic has hired the renowned law firm Wilson Sonsini to actively prepare for an IPO plan, potentially launching as early as 2026, targeting a valuation exceeding $300 billion, with investors including Microsoft and Nvidia. The aim is to go public before OpenAI and seize pricing initiative. OpenAI has also begun IPO preparations, seeking a $1 trillion valuation. The listing processes and market performance of both will directly test the sustainability of the 'AI bubble' and high valuations.

Google Workspace Studio is now available, enabling all enterprise users (including Business and Enterprise tiers) to build zero-code, proprietary AI agents powered by Gemini 3. This allows for cross-system workflow automation across Gmail, Drive, Docs, Chat, and third-party applications (e.g., Salesforce, Jira). Supporting multimodal task understanding, modular steps, API extensions, and custom triggers, it strengthens underlying office automation capabilities, advancing the vision of 'everyone as an AI workflow architect' in enterprises.

Global financial regulators like the Bank of England warn that driven by massive AI infrastructure financing, tech stock valuations are nearing pre-2008 financial crisis extremes. The market is highly interconnected with tight stock-bond linkages, meaning a correction in the AI sector could trigger a chain reaction of credit risks. The Fed and IMF are also monitoring whether this wave of AI investment will replay the 'dot-com bubble', but note that leading AI firms are mostly profitable, not purely speculative, indicating differences in risk structure.

NVIDIA's new GB200 Blackwell NVL72 servers, designed for Mixture-of-Experts (MoE) model inference (e.g., Kimi K2, DeepSeek), deliver 10x the performance of H200. Leveraging chip-level co-design, distributed attention, and decoupled inference, supply has commenced for top clients like Microsoft and Amazon, as well as the open-source model community. This will drive practical commercial deployment of ultra-large models and the development of multi-agent autonomous systems.

Amazon, in partnership with Visa and others, has launched agentic chatbot tools and commercial payment APIs for developers, enabling AI agents to autonomously complete product searches, ordering, payment, and B2B/retail settlement. Concurrently, the Agentic Commerce Protocol has gained participation from major players like Stripe, advancing the standardization of infrastructure for AI 'autonomous consumption' capabilities. Companies like Google and Microsoft are also widely integrating agents (AI agents) into existing mainstream systems for workflows and business scenarios, signaling that Agentic AI has become the core engine of the new generation of productivity.

An internal training guidance file (Soul Document) for the Claude 4.5 Opus model was successfully extracted and made public by an external engineer. The document details the model's priority principles, ethical boundaries, behavioral baselines, and 'functional empathy' settings. Anthropic has confirmed its authenticity and role in model fine-tuning, systematically revealing how cutting-edge large model vendors embed ethical norms, role perception, and identity into AI agents, pushing a new direction for AI ethics engineering.

A survey of 132 internal engineers by Anthropic reveals that over 60% of work tasks are now assisted by Claude, with overall engineer productivity increasing by 50%. The new agent toolchain supports up to 20-step automated workflows. However, employees widely express anxiety about skill erosion, job security, and guidance mechanisms for junior engineers. This reflects that while AI-assisted development accelerates industry model upgrades, it also challenges the stability of engineering education and talent structures.

At its re:Invent conference, Amazon officially launched multiple Nova large models, covering text, image, and video generation, with overall costs 75% lower than competitors. They will first satisfy internal e-commerce/cloud service needs, with APIs opening after thorough testing. Amazon also announced building Rainier, the world's largest AI compute cluster, and pledged to deepen its partnership with Anthropic (with an additional $40 billion investment), aiming to solidify its dominance in the cloud services AI ecosystem.

Meta announced poaching Apple's User Interface (UI) design head, Alan Dye, at a high salary to lead a new creative design studio within its Reality Labs. He will oversee integrated R&D for AI hardware, software, and human-machine interaction experiences. Meta also recruited several key Apple design personnel, highlighting the worsening brain drain of core Apple executives. This foreshadows escalating competition in forms like AR/MR and AI glasses. The industry is watching whether design innovation and experience differentiation will enter a new phase.

Google is trialing AI-generated news headlines and summaries in its mobile Discover feed. Some rewritten headlines exhibited inaccuracies, misleading information, or even conflict with the original article, drawing strong criticism from publishers and news coalitions questioning its secondary creation mechanism. This experiment coincides with Google's full-scale push to integrate its AI Mode and AI Overview features, highlighting the risks of large model content generation and the urgent need for new standards regarding text traceability.