Three engineers conducted an experiment putting large language models—GPT, Claude, and Grok—in control of a real Toyota Corolla to test real-world autonomous driving capabilities. The test was designed to see if frontier LLMs could navigate unscripted driving tasks, such as getting to an In-N-Out restaurant without explicit directions. Only one of the three models successfully completed the task, highlighting both the potential and limitations of using general-purpose language models for specialized physical-world tasks.
Anthropic has announced an expansion of its Claude Startup program, a competitive initiative aimed at attracting founders and rapidly growing companies to build products on Claude, Anthropic's AI model. The program provides companies with discounted API access, technical support, and co-marketing opportunities. By expanding the program, Anthropic is attempting to build an ecosystem of applications around Claude—similar to what OpenAI has done with ChatGPT.
Meta's AI assistant Muse has arrived on iPad just one month after its mobile launch, signaling the company's rapid push to embed its AI agent across devices and platforms. Muse was designed as a conversational AI that can help users with a variety of tasks—from information lookup to content creation. The iPad version includes tablet-optimized features and integrations that take advantage of the larger screen.
Microsoft demonstrated a significant upgrade to Copilot that grants the AI assistant direct access to local Windows files and the ability to execute actions across the operating system. Under Microsoft's "Hybrid Intelligence" framework, Copilot can now search your personal documents, launch applications, modify settings, and execute multi-step workflows autonomously. This shift moves Copilot from a passive information tool to an active agent capable of managing system-level tasks without user intervention for each step.
Microsoft announced the Surface Laptop Ultra, a new flagship laptop starting at $2,599 that runs Nvidia's RTX Spark Arm-based chip. The machine is designed specifically to run AI models and agents locally on the device, eliminating reliance on cloud services for certain tasks. Alongside the hardware, Microsoft revealed a significant upgrade to Copilot that gives the AI assistant direct access to local files and the ability to execute actions across Windows itself.
Nous Research, a developer of specialized AI models and agents, confirmed it has reached a $1.5 billion valuation after closing a $90 million Series B funding round. The company is best known for its Hermes models and agents, which are optimized for specific business tasks. With this capital injection, Nous is launching AI agents designed specifically for business users—moving beyond open-source models into application-layer products.
OpenAI has introduced College Planner to ChatGPT for Teens, a feature designed to help high school students manage the college application process. The tool works within ChatGPT to provide guidance on essay writing, deadline management, and school comparison. Alongside College Planner, OpenAI is rolling out flashcards and quiz functionality—study aids that allow students to create and test themselves on custom content. The company has also established a Teen AI Council, giving younger users a voice in how ChatGPT's teen-focused features develop.
OpenAI has released GPT-6 globally with a new feature called Intelligent UI that transforms how ChatGPT presents information. Instead of text-only responses, the chatbot can now generate interactive charts, diagrams, forms, and tappable buttons alongside its written answers. This visual layer allows users to explore data and interact with content directly within the chat interface.
The U.S. Department of Defense has launched Tradewinds, an initiative designed to streamline procurement processes and make it faster for military agencies to deploy AI systems from both traditional defense contractors and non-traditional tech companies. Through Tradewinds, companies like OpenAI, Anthropic, and Google can pitch AI solutions directly to the Pentagon with simplified approval processes. The initiative replaces lengthy acquisition cycles with shorter, video-based pitches—sometimes as brief as five minutes—to help agencies evaluate and deploy AI more rapidly.
Radisson Hotel Group, in partnership with Accenture, has built a ChatGPT plugin that lets travelers discover, compare, and book hotels directly within the chat interface. Instead of switching between ChatGPT and a hotel booking site, users can describe their travel needs to ChatGPT, receive personalized recommendations from Radisson's inventory, and complete reservations without leaving the conversation. Accenture built the integration using OpenAI's APIs, demonstrating how established enterprises can rapidly deploy AI-native booking experiences.