AI Native Weekly Newsletter: 05 July 2025
Elsewhere, Meta recruits nearly 20 elite researchers from OpenAI, Google, and Anthropic for its superintelligence team. Oracle commits 4.5GW computing power to OpenAI’s Stargate project, while Microsoft open-sources GitHub Copilot Chat and Anthropic experiments with autonomous AI store management. From talent acquisition to infrastructure scaling, this week’s developments show AI capabilities becoming more strategic, autonomous, and market-defining than ever.
Contents
-
Why AI Native Companies Are Winning: From Efficiency to Opportunity
-
Meta’s AI Talent War: Zuckerberg Unveils Superintelligence Team
-
Oracle-OpenAI Stargate Deal Expands with 4.5GW Data Centers
-
Anthropic Tests Claude as Autonomous Store Manager
-
Cursor Web & Mobile: Code with AI Agents Anywhere
-
VS Code Open Sources GitHub Copilot Chat Extension
-
Qwen VLo: From Understanding to Creating Visual Content
-
Tencent Launches Hunyuan-A13B: Open-Source AI Model
Why AI Native Companies Are Winning: From Efficiency to Opportunity
Traditional companies pursuing “Efficiency AI” for 10-50% productivity gains risk being displaced by AI Native startups building entirely new business models. This analysis reveals why incumbents optimizing existing workflows get trapped in legacy systems, while AI Native competitors design autonomous systems that act on future signals rather than react to past data. The key strategic shift: stop asking “How can AI make us faster?” and start asking “What can we do now that was previously impossible?” Companies that embrace Opportunity AI will capture the most valuable markets while efficiency-focused incumbents get commoditized.
Meta’s AI Talent War: Zuckerberg Unveils Superintelligence Team
Mark Zuckerberg revealed Meta’s new superintelligence team in an internal memo, featuring nearly 20 elite researchers poached from OpenAI, Google, and Anthropic. The hiring spree follows Meta’s $14.3B investment in Scale AI and appointment of Alexandr Wang as chief AI officer. Notable hires include creators of GPT-4o, o3 models, and Gemini systems, signaling Meta’s aggressive push in the AI talent war through its new Meta Superintelligence Labs.
Oracle-OpenAI Stargate Deal Expands with 4.5GW Data Centers
OpenAI will rent 4.5 gigawatts of computing power from Oracle for the Stargate AI initiative – enough to power millions of homes. This forms part of Oracle’s $30 billion annual deal starting fiscal 2028. Oracle will develop data centers across Texas, Michigan, Wisconsin, Wyoming, and other states. The partnership supports the $500 billion Stargate project. Oracle shares jumped 5% to record highs.
Anthropic Tests Claude as Autonomous Store Manager
Anthropic had Claude operate a small automated store in their office for a month. The AI “Claudius” managed inventory and pricing but failed financially due to poor decisions: selling at losses, offering excessive discounts, ignoring profitable opportunities. Despite failures, the experiment suggests AI business management is promising with improved tools. Claude also experienced identity confusion before recovering.
Cursor Web & Mobile: Code with AI Agents Anywhere
Cursor Agents now work on web and mobile browsers, enabling remote coding tasks like bug fixes, feature development, and codebase analysis. Install as PWA on iOS/Android via browser menu. Features include Slack integration with @Cursor triggers, team collaboration with PR creation, parallel agent execution, and rich context support with images. Seamlessly sync work back to desktop IDE for review and editing.
VS Code Open Sources GitHub Copilot Chat Extension
Microsoft reaches first milestone in its open source AI editor initiative by releasing GitHub Copilot Chat extension code under MIT license. This marks the beginning of VS Code’s transition to fully open source AI capabilities, with plans to integrate inline completions and refactor components into VS Code core over the coming months. The move enables community innovation and full transparency into AI implementation details.
Qwen VLo: From Understanding to Creating Visual Content
Alibaba released Qwen VLo, a multimodal AI model that understands and generates images. It features progressive generation, supports natural language editing commands, multiple languages, and dynamic aspect ratios. The model can perform image generation, style transfer, object detection, and complex multi-task operations. Available on
Tencent Launches Hunyuan-A13B: Open-Source AI Model
Chinese tech giant Tencent releases Hunyuan-A13B, an open-source AI model using MoE (Mixture of Experts) architecture that selectively activates relevant components for efficiency. With 80B total parameters and 13B active parameters, it delivers top-tier performance while drastically reducing computational costs. The model can run on just one mid-range GPU, making advanced AI accessible to individual developers and small businesses through GitHub and HuggingFace.