Good morning. It’s Friday, June 13th.

In today’s email:

• NVIDIA Goes All-In on Robotics

• The New AI Browser

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

Nvidia Bets on Physical AI as $50 Trillion Industrial Backbone

Physical AI is fast becoming Nvidia’s defining bet on the future of industry, intelligence, and infrastructure. At recent events in Paris, CEO Jensen Huang laid out a sweeping vision that positions robotics, autonomous systems, and industrial AI as the backbone of a $50 trillion market opportunity. This effort is already materializing through Nvidia’s Drive AV platform, now live in the Mercedes CLA and set to roll out across Volvo and JLR vehicles, with new Cosmos models enhancing performance in challenging environments.

To scale AI-driven manufacturing, Nvidia is building its first industrial AI cloud in Germany, a 10,000-GPU hub using DGX B200 and RTX PRO servers, while expanding tech centers and committing to 20 AI factories across Europe. Sovereign AI is now strategic infrastructure, with DGX Cloud Lepton and national providers offering over 3,000 Blackwell-based exaflops to enterprises. Partnerships with Mistral and Novo Nordisk highlight the platform’s breadth, from advanced AI deployment to accelerated drug discovery via the Gefion supercomputer.

Supporting this push into embodied intelligence, Nvidia also open-sourced Gr00t N1.5, a 3B-parameter foundation model designed for humanoid reasoning and compatible with Hugging Face’s LeRobot SO-101 Arm. Read more.

Dia Is a Chat-First Browser That Knows What You’re Doing

Dia is a new AI-powered web browser from The Browser Company designed to act more like an intelligent assistant than a traditional browser. At the center of its interface is a built-in chatbot that stays with you as you browse, able to reference your open tabs, past activity, and even logged-in accounts to help you search, summarize, answer questions, or complete tasks, all without switching between apps or tabs.

What makes Dia unique is its ability to understand what you’re trying to do and route your request to the right AI tool, whether it’s for shopping, writing, or research. The address bar doubles as a command line, where you can type anything, from a search to a task. With access to local cookies and browsing data, Dia builds contextual memory that evolves with use, while keeping sensitive data encrypted and mostly device-local. Instead of trying to outdo other AI models, it focuses on seamlessly combining them to make browsing smarter and more fluid. Read more.

• Meta Assembles AGI Stack, Buys a Data Future with $15B Scale AI Deal

• Microsoft-backed Mistral launches European AI cloud to compete with AWS and Azure

• OpenAI Strengthens Industry Links with AMD Chips and Mattel’s Intelligent Play

• Vision Pro Powers Real-Time Robot Control via Closed-Loop AI

• xAI tests Voice Mode and Tasks on Grok web with signs of Grok 3.5 integration

• DeepMind Takes on Hurricanes, Google Installs New AI Chief During Cloud Turmoil

• Here’s the $2,000 fully AI-generated ad that aired during the NBA Finals

• Anthropic released "AI Fluency" a free online course to Learn to collaborate with AI

• NVIDIA Boosts AI Image Generation: 2x Faster Stable Diffusion 3.5, 40% Lower Memory Demand

• Vibe Coding Is Coming for Engineering Jobs

• Command Line Meets Agentic AI as Atlassian Rolls Out Rovo Dev to Developers

• Nvidia’s Jensen Huang says he disagrees with almost everything Anthropic CEO Dario Amodei says

• Microsoft’s new Copilot Vision can ‘see’ your apps on Windows

• Disney and NBC Universal take 'copyright free-rider' Midjourney to court over alleged plagiarism

• Microsoft just announced that Copilot Vision is now free to try on mobile

• Swedish fighter jet autonomously flies real combat mission using onboard AI

• Starbucks' new game plan to roll out AI chatbots at cafes could serve as a 'litmus test' for the industry, analyst says

• Astra by Topaz Labs Upscales AI-Generated Videos to 4K Using Starlight Models

• Apple is salvaging Image Playground with a boost from ChatGPT

• Training robots without robots: Smart glasses capture first-person task demos

📄 Confidence Is All You Need: Few-Shot RL Fine-Tuning of Language Models

📄 Multiverse: Your Language Models Secretly Decide How to Parallelize and Merge Generation ​​

📄 Text-Aware Image Restoration with Diffusion Models

📄 PlayerOne: Egocentric World Simulator

📄 Autoregressive Semantic Visual Reconstruction Helps VLMs Understand Better  

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on 𝕏!