Good morning. It’s Wednesday, February 19th.

On this day in tech history: 1946, Alan Turing presented his "Proposal for the Development in the Mathematics Division of an Automatic Computing Engine (ACE)" to the National Physical Laboratory (NPL) in England.

In today’s email:

• Grok 3 Leads The Pack

• New Humanoid “Mentee” Robot

• Former OpenAI CTO Starts New Company

• 5 New AI Tools

• Latest AI Research Papers

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

Grok-3 outpaces ChatGPT, DeepSeek and becomes the new king of the hill

Elon Musk’s xAI debuted Grok-3 on Monday and the Chatbot Arena community has unanimously declared Grok-3 the new benchmark leader, with standout results in AIME and GPQA. Blind tests on the platform placed Grok-3 ahead of ChatGPT, Gemini, and DeepSeek surpassing the 1400 score barrier on LMArena, excelling in coding, mathematics, and instruction following. OpenAI founding member Andrej Karpathy shared an early assessment on 𝕏, praising Grok-3’s performance and suggesting it matches or exceeds OpenAI's top-tier models.

The model runs on xAI’s Colossus supercomputer, which has now expanded to 200,000 Nvidia GPUs—doubling its previous capacity in just 92 days. During the live demo, Musk confirmed that Grok 2 will be open-sourced once Grok-3 is stable, likely within the next few months. He also hinted that Grok might be tagging along Optimus robots to Mars, with a potential launch window opening in November 2026.

Premium+ 𝕏 subscribers get first dibs on features like DeepSearch and an upcoming voice mode designed to enhance functionality. A separate web and app subscription will follow. Premium users get first dibs on features like DeepSearch and a voice mode set to enhance functionality. @minchoi’s thread presents some wild use cases. Read more.

Israeli company introduces humanoid robot with 360° vision, enhanced power, and precision

Mentee Robotics from Israel has unveiled the MenteeBot V3.0, an advanced humanoid robot designed for both domestic and industrial applications. This model features custom-built actuators that offer three times the power, precision, and efficiency compared to its competitors. With a height of 175 cm and a walking speed of 1.5 meters per second, it can lift up to 25 kg.

The robot's 360-degree vision system, powered by head-mounted sensors and fisheye cameras, provides comprehensive environmental awareness. A hot-swappable battery system allows for over three hours of continuous operation, enhancing its utility in demanding environments.

The MenteeBot V3.0 is equipped with dual Jetson Orin AGX processors for advanced AI functions and real-time decision-making. It employs a unique training method called the "Foundation Model," combining text instructions and live demonstrations for effective learning. Read more.

Former OpenAI CTO Mira Murati's new AI startup, Thinking Machines Lab, comes out of stealth

Mira Murati has officially unveiled Thinking Machines Lab, a new AI venture stacked with top-tier talent from her previous team. The announcement, made in a blog post, comes months after her departure from OpenAI, where she played a central role in shaping its AI research and deployment strategies.

The startup’s focus on “frontier” AI signals a push into high-complexity, next-generation models—potentially addressing core gaps in scalability, alignment, and safety. Among those joining her is OpenAI cofounder John Schulman, who briefly worked at Anthropic before linking up with Murati. Thinking Machines Lab is expected to contribute to the broader discourse on AI safety, governance, and technical advancements, though its exact direction remains to be seen. Read more.

• DeepSeek NSA claims 11x faster processing for large datasets

• China’s Tencent says its trialling own AI reasoning model alongside DeepSeek to make responses more ‘human’

• Ilya Sutskever’s AI Startup Reportedly Raising $1B+ at $30B Valuation

• AI agents prone to basic manipulation attacks, study says

• Perplexity AI Open-Sources R1 1776 to Remove Censorship from DeepSeek-R1

• AI can fix bugs—but can’t find them: OpenAI’s study highlights limits of LLMs in software engineering

• Anthropic prepares to launch Thinking Mode and Web Search for Claude

• ​​OpenAI CEO Sam Altman all but announces return to open source via 𝕏 survey

• Microsoft AI releases OmniParser V2: an AI tool that turns any LLM into a computer use agent

• Replit and Anthropic’s AI just helped Zillow build production software—without a single engineer

• Mistral releases regional model focused on Arabic language and culture

• Chinese AI startup Skyworks debuted a new open video AI: Skyreels-V1

• LLMs are now capable to model cognitive processes of fruit flies, rats, and humans. They model us.

• US Navy uses AI to train laser weapons against drones

• AI 'brain decoder' can read a person's thoughts with just a quick brain scan and almost no training

• AI Tools shift knowledge workers from problem solvers to verifiers, Microsoft study 

• Humane is shutting down the AI Pin and selling its remnants to HP

• OpenAI board considers special voting powers to prevent Elon Musk takeover

• Meta announces LlamaCon, its first generative AI dev conference

• Google brings Deep Research AI tool to Gemini mobile app

• Microsoft launches AI Hub in Windows Store to showcase AI experiences

• What the US' first major AI copyright ruling might mean for IP law

• DarkMind: A new backdoor attack that leverages the reasoning capabilities of LLMs

• Scientists created the lightest and strongest nanomaterial ever

📄 Large Language Diffusion Models

📄 The Danger of Overthinking: Examining the Reasoning-Action Dilemma in Agentic Tasks ​​

📄 Learning Getting-Up Policies for Real-World Humanoid Robots

📄 Soundwave: Less is More for Speech-Text Alignment in LLMs

📄 Continuous Diffusion Model for Language Modeling

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!