Good morning. It’s Wednesday, July 9th.

On this day in tech history: In 1956, the Dartmouth Summer Research Project on Artificial Intelligence was in full swing, where John McCarthy, Marvin Minsky, Claude Shannon, and Nathaniel Rochester proposed that “every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it.” By July 9, they were deeply engaged in foundational discussions on symbolic reasoning, neural networks, and automata theory, laying the groundwork for what would become formal AI research.

In today’s email:

• Grok 4 Today

• OpenAI Poaches Engineers

• HuggingFace Robot

• Meta Pulls Apple’s Head of AI

• Robots With Inner Monologue

• 5 New AI Tools

• Latest AI Research Papers

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

Be the Person on Your Team Who Knows How to Actually Use AI

The AI for Business & Finance Certificate Program from Columbia Business School Exec Ed & Wall Street Prep delivers hands-on experience directly from from AI leaders and Columbia’s world-class faculty and guest speakers

Gain the skills that will turn you into an “AI first” leader in your role.  

• Skip the theory and go straight to practice by tackling real use cases currently being implemented at the world’s most AI-forward financial institutions and corporations

• Learn from Columbia Business School faculty and directly from those building AI tools at BlackRock, Morgan Stanley, Ramp, Perplexity, and OpenAI

• Join a rigorous program with no coding or tech background needed—we'll guide you step by step from fundamentals to real-world application

✅ Earn a certificate from Columbia Business School Exec Ed in just 8 weeks—and gain lifetime access to course materials and a global professional network.

Save $300 using code AIBREAKFAST when you enroll today + an extra $200 on early enrollment.

Learn More

Grok 4 launches today as xAI expands its model lineup

xAI rolls out Grok 4 today alongside Grok Code 4 and an “Extended” variant buried in upsell code. Rumors tie “Extended” to the long-awaited Big Brain feature for deeper reasoning. Leaked benchmarks peg Grok 4 at 45% on the Humanity Last Exam, up from Grok 3, targeting devs and technical users needing stronger output. Voice mode and screen sharing may drop too, adding real-time collaboration muscle.

But Grok’s rollout could stall after its bot was silenced on 𝕏 for unfiltered content, throwing fresh light on xAI’s moderation gaps. Read more.

OpenAI Poaches 4 High-Ranking Engineers From Tesla, xAI, and Meta

OpenAI just pulled four top engineers from Tesla, xAI, and Meta to bulk up its scaling team, per an internal Slack memo from cofounder Greg Brockman. David Lau, Tesla’s former VP of software, joins along with Uday Ruddarraju and Mike Dalton, ex-xAI engineers who helped build Colossus, the 200,000-GPU supercomputer, and Angela Fan, an AI researcher from Meta. This crew will reinforce Stargate, OpenAI’s massive backend system powering its next-gen foundation models.

Speaking at Sun Valley, CEO Sam Altman didn’t hold back on the escalating talent fight with Meta, dropping a vague promise that OpenAI’s upcoming hardware “will be great.” He also took a swipe at smart glasses hype, despite his bold frames, he says they’re just glasses. Days earlier, Altman broke ranks with Democrats, calling himself “politically homeless” and blasting what he sees as anti-innovation, anti-billionaire politics choking Silicon Valley’s edge. Read more.

Hugging Face opens up orders for its Reachy Mini desktop robots

Hugging Face has dropped Reachy Mini, a $299 desktop robot built to crack open the robotics industry. At just 11 inches tall, this DIY humanoid hooks straight into the Hugging Face Hub, giving millions of developers plug-and-play access to AI models. Reachy Mini runs on a Raspberry Pi 5 with full wireless autonomy, six degrees of freedom, a camera, mics, and animated antennas, packing serious capability for its price.

Its newly launched SmolLM3 squeezes multilingual, long-context reasoning into just 3 billion parameters, handling up to 128k tokens with streamlined linear and grouped attention. Trained on 11T tokens, it rivals heavyweight 7B+ models for tool use, few-shot tasks, and multilingual QA, but runs light enough for edge hardware and private stacks. Read more.

Meta reportedly recruits Apple's head of AI models

Meta has reportedly hired Ruoming Pang, Apple’s head of AI models, to join its advanced AI superintelligence unit, Bloomberg revealed. Pang led Apple’s foundational AI model development, including the systems powering Apple Intelligence and on-device features. His exit signals growing instability within Apple’s AI division, which has struggled to match rivals like OpenAI and Anthropic.

Insiders suggest more departures could follow. With Siri still underperforming, Apple is reportedly looking to plug the gap with third-party models. For Meta, it’s another high-profile win in a growing stack of AI hires. Pang brings expertise in lightweight, on-device AI, exactly the kind of talent Meta CEO Mark Zuckerberg is stacking up from DeepMind, OpenAI, and Safe Superintelligence. Read more.

DeepMind Files Patent for AI Robots with Inner Monologue

DeepMind has filed a patent for a system enabling AI agents and robots to generate “inner speech” to better understand and learn tasks. The technology uses image or video input, such as a person picking up a cup, and pairs it with natural language descriptions via a language model. This internal monologue helps the agent link visual input with contextual reasoning, improving decision-making.

The approach enables zero-shot learning, allowing robots to handle unfamiliar objects without prior training. DeepMind claims this method boosts learning efficiency while reducing memory and compute needs. It follows last week’s release of DeepMind’s compact, on-device vision-language robotics model. As AI robotics evolve, adding internal reasoning is seen as key to reducing unpredictability. Google, Nvidia, and Intel are all pursuing similar patents as competition in robotics AI intensifies. Read more.

• Build Scalable AI Workflows Using BeeAI: A Practical Tutorial

• Study: Innovation, Not Efficiency Drives AI ROI

• World’s first garage-built humanoid robot open to creators worldwide

• AI is forcing the data industry to consolidate — but that's not the whole story

• Microsoft Open-Sources GitHub Copilot Chat Extension for VS Code—Now Free for All Developers

• NVIDIA’s Highly Anticipated “Mini-Supercomputer,” the DGX Spark, Launches This Month

• ChatGPT is testing a mysterious new feature called 'study together'

• A developer focused on stopping AI bots says poisoning datasets is like peeing in the ocean

• AI war erupts as Huawei accused of stealing Alibaba model in major China tech clash

• Russia's new drone uses Nvidia's mini supercomputer to make its own kill list

• Anthropic proposes targeted transparency framework for frontier AI systems

• Watch: Proactor v1.0, world’s first ‘self-active’ AI teammate

• Cursor apologizes for unclear pricing changes that upset users

• Researchers reveal that AI models have distinct strategic fingerprints in classic game theory tests

• Moonvalley’s ‘ethical’ AI video model for filmmakers is now publicly available

• NVIDIA Makes Massive AI Prompts Fast and Scalable for 32× More Users

• Dynamics Lab launches Mirage, an AI-powered game engine for user-generated content

• xAI’s Grok Models are Now on Oracle Cloud Infrastructure

• LangChain is about to become a unicorn, sources say

• Teachers union partners with Anthropic, Microsoft and OpenAI to launch AI-training academy

• In a blow to Google Cloud, Replit partners with Microsoft

• Scientists burned, poked and sliced their way through new robotic skin that can 'feel everything'

• Black Panther: China’s robot dog breaks Boston Dynamics speed record

• The world is too complex for AI to pick your stocks, a hedge fund quant says

📄 4DSloMo: 4D Reconstruction for High Speed Scene with Asynchronous Capture

📄 DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge ​​

📄 MemOS: A Memory OS for AI System 

📄 Agent KB: Leveraging Cross-Domain Experience for Agentic Problem Solving

📄 RLVER: Reinforcement Learning with Verifiable Emotion Rewards for Empathetic Agents

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!