The morning briefing.
While you slept: Anthropic's AI is writing 65% of its own code, China forecasts 50,000 humanoid robot shipments, and Europe pushes controversial "Chat Control" legislation.
AI Agent Adoption. Anthropic's internal use of Claude Tag for Slack is generating 65% of its product team's code, showcasing significant enterprise AI agent adoption. This integration allows AI to proactively assist in workflows, moving beyond simple queries. Tencent is also launching Agently Mail, a dedicated email service for AI agents, enabling autonomous email and business coordination. These developments highlight a growing trend towards embedding AI agents directly into operational processes for enhanced productivity.
Frontier Model Development. The Tsinghua ecosystem, including Zhipu AI and Shengshu Tech, is focusing on world models as the next AI frontier, spanning video, robotics, and autonomous driving. Meanwhile, Moonshot AI's Kimi K2.7 Code introduces a new paradigm in AI coding, shifting from code generation to reconstructing behavior from existing products. Mistral AI has also released OCR 4, which reportedly outperforms competitors in 72% of blind tests for document text extraction. These efforts push the boundaries of AI capabilities and efficiency.
AI Security and Costs. Anthropic's Mythos model identified vulnerabilities in classified US government systems, underscoring AI's potential in cybersecurity but also the risks involved. Simultaneously, database vendors like Pinecone and Tiger Data are positioning themselves as solutions to runaway AI costs, advocating for smarter data plumbing to reduce token use. Austrian startup Ora Computing recently raised €3.5 million to develop software that optimizes and compresses AI foundation models, further addressing efficiency concerns. These initiatives reflect a dual focus on securing AI systems and making them more economically viable.
Geopolitical Tech Shifts. Morgan Stanley has significantly increased its forecast for China's humanoid robot shipments to 50,000 units this year, driven by commercial validation and policy support. This comes as China Telecom awarded a US$1.7 billion server procurement deal to Huawei partners, reinforcing domestic hardware reliance. On the policy front, the European Commission's President Metsola controversially overrode MEPs to push through "Chat Control" legislation, raising privacy concerns. These events highlight ongoing geopolitical competition and regulatory challenges in the tech sector.
Anthropic's Claude Tag AI now writes 65% of internal code in Slack
Anthropic reports that its Claude Tag AI, embedded in Slack, now generates 65% of the code for its product team. This tool allows teams to assign tasks to @Claude in any channel, demonstrating significant internal AI agent adoption.
Morgan Stanley raises China humanoid robot shipment forecast to 50,000 units
Morgan Stanley has increased its forecast for China's humanoid robot shipments this year to 50,000 units from 28,000. This upgrade reflects accelerating commercial validation, policy support, and supply-chain momentum.
Huawei partners win big as China Telecom invests US$1.7 billion in 40,000 servers
China Telecom awarded a US$1.7 billion high-performance server procurement deal to domestic CPU chip companies, including Huawei Technologies. This move supports state firms in replacing US technology with domestic hardware for 40,000 servers.
European Commission's Metsola overrides MEPs to force through Chat Control
European Commission President Roberta Metsola has overridden MEPs to push through the controversial "Chat Control" legislation. This action forces a child abuse law despite parliamentary opposition.
Trump administration asks Meta to allow vetting of its AI models
The Trump administration has requested Meta to allow vetting of its AI models, though the executive order does not legally require compliance. This highlights ongoing discussions around AI oversight.
Mistral's new OCR model beats competitors in 72% of blind test cases
Mistral AI has released OCR 4, a new model designed to read text from documents like PDFs and Word files. The company states that OCR 4 outperforms competitors in 72% of blind test cases.
Anthropic's Mythos model found vulnerabilities in classified US government systems
Anthropic's Mythos model has reportedly identified vulnerabilities within classified US government systems. It remains unclear if these vulnerabilities could be immediately exploited.
Tencent QQ Mail launches Agently Mail for autonomous AI agent coordination
Tencent has launched Agently Mail, a dedicated email service in beta testing, specifically designed for AI agents. This service aims to enable autonomous email and business coordination for AI systems.
Kimi K2.7 Code introduces new paradigm in AI coding: reconstructing behavior
Moonshot AI's Kimi K2.7 Code is shifting the approach to AI coding from generating code to reconstructing behavior based on existing products. This represents an overlooked new paradigm in the field.
Tsinghua ecosystem targets world models as the next AI frontier
The Tsinghua ecosystem, including companies like Zhipu AI and Shengshu Tech, is focusing on world models as the next frontier in AI. Their efforts span applications in video, robotics, and autonomous driving.
Database vendors pitch smarter data plumbing to cure runaway AI costs
Database vendors Pinecone and Tiger Data are promoting their solutions as a way to manage escalating AI costs. They suggest that smarter data plumbing can significantly cut token use and optimize agentic workloads.
Ora Computing raises €3.5M to build efficiency layer of the AI stack
Ora Computing, a startup focused on optimizing and compressing AI foundation models, has closed a €3.5 million seed funding round. The investment was led by Constructor Capital and Greencode Ventures.
MoEngage acquires US-based AI startup Aampe to scale agentic marketing
SaaS major MoEngage has acquired San Francisco-based AI infrastructure startup Aampe to enhance its customer engagement stack. This acquisition aims to scale agentic marketing capabilities.
QOSMIC bags $3.3M to build laser-based satellite communication systems
Spacetech startup QOSMIC has raised $3.3 million in a seed funding round led by Accel and Prosus. The funds will be used to develop laser-based satellite communication systems.