🎬 OpenAI kills Sora, redirects compute to 'Spud' 🎬
OpenAI's replacement strategy sacrifices its flagship video tool to power a bigger AI play.

OpenAI has fully shut down Sora — its app, API, and all video features inside ChatGPT are now gone. The company is reallocating the freed compute to develop its next model, codenamed 'Spud,' which CEO Sam Altman reportedly told staff could "really accelerate the economy." The entire Sora team is being reassigned to robotics and "world simulation" initiatives, a signal that compute allocation now drives OpenAI's product strategy. With Sora gone, Google becomes one of the only major players still pushing AI-generated video at scale. OpenAI clearly sees greater long-term value in foundational model capability than in the increasingly crowded generative video space.

🤖 Claude Auto Mode lets agents run without approval 🤖
Anthropic's research preview rollout lets Claude act independently across multi-step tasks.

Anthropic has released Claude Auto Mode in research preview, a capability that allows Claude to autonomously execute multi-step actions without requiring user approval at each stage. The feature includes built-in safeguards to filter risky behavior and defend against prompt injection attacks. Auto Mode positions Claude as a more capable autonomous assistant for enterprise and developer workflows — a direct competitive push in the race to build reliable AI agents. The research preview status means Anthropic is collecting real-world feedback before a broader rollout. With enterprise adoption of agentic tools accelerating, this could reshape how teams handle complex, multi-step processes while maintaining safety guardrails.

⚡ Google's TurboQuant cuts AI memory costs ⚡
Google's new compression technique could cut AI inference pricing across the industry.

Google Research has published TurboQuant, a quantization method that sharply reduces vector memory overhead while preserving model performance. The technique targets key-value cache efficiency — a critical bottleneck in large language model inference — and also accelerates vector search operations. By achieving extreme compression without meaningful quality loss, TurboQuant could lower infrastructure costs for both cloud-based and on-device AI deployment, directly affecting competitive pricing. The method addresses one of the most pressing challenges in scaling production AI systems: keeping serving costs manageable as models grow larger. This signals Google's continued investment in efficiency research, an area that underpins its ability to offer AI services at scale.

INTERESTING TO KNOW

💰 Anthropic hits $19B ARR as Apple runs Claude internally 💰

Anthropic is now generating $19 billion in annualized revenue, a 20x increase in roughly 15 months, driven largely by enterprise adoption and Claude Code, which accounts for about 4% of all public GitHub commits. Meanwhile, Bloomberg reports that Apple internally "runs its business on Anthropic" for product development — hosting Claude on its own servers for code generation, prototype testing, and security automation — even as it publicly partners with Google Gemini for Siri. The enterprise concentration is notable: eight of the ten largest Fortune 10 companies are Anthropic customers, signaling where large-scale AI spend is shifting.

🦾 Amazon buys Fauna Robotics for humanoid push 🦾

Amazon has acquired Fauna Robotics, the maker of human-like service robots, in a move to integrate humanoid capabilities into its logistics and retail operations. The acquisition extends Amazon's robotics ambitions beyond warehouse automation into general-purpose, human-form-factor machines. The deal intensifies competitive pressure in the humanoid robotics sector, where Tesla, Figure AI, and Unitree are all advancing rapidly. Financial terms were not disclosed.