A humanoid robot has just done something unusual. Instead of walking around a lab or a warehouse, it headed out into deep snow and freezing weather and kept going for hours.

The robot, called G1 and built by Chinese robotics company Unitree, walked across a snowfield in northwest China where temperatures dropped to about minus 47 degrees Celsius. That is colder than conditions most machines are designed to handle, yet it managed to keep moving without stopping.

During the test, the robot took more than 130,000 steps on its own, making it the first time a humanoid robot has carried out long autonomous walking in such extreme cold. The journey was not random. Its path was carefully planned so that its footprints formed a giant symbol in the snow measuring about 186 metres by 100 metres, large enough to be seen from above.

To survive the conditions, engineers prepared it in practical ways. The robot wore an insulated jacket and had protective coverings added around its joints and batteries to stop them freezing. Navigation relied on satellite positioning along with onboard sensors that helped it plan each step and avoid slipping on uneven ground.

Humanoid robots usually operate indoors or in controlled environments, so this kind of test matters. Cold weather can affect electronics, power systems, and moving parts. Showing that a robot can function outside in harsh conditions suggests these machines may eventually be used in more real world settings such as outdoor inspection, exploration, or rescue support.

The walk itself may look like a demonstration, but it highlights how robotics is slowly moving beyond staged environments. For now, it is another sign that machines built to resemble human movement are being pushed into places where people normally go, including some of the coldest ones.

OpenAI released a new coding model called GPT-5.3 Codex only minutes after Anthropic introduced its own update, Claude Opus 4.6. The timing was not accidental. Both companies had planned to announce their tools at the same time, but Anthropic shifted its release forward by about fifteen minutes to go first.

OpenAI describes GPT-5.3 Codex as its most capable coding model so far. The system is designed to act more like a working assistant than a simple code generator, handling research tasks, debugging, and deployment workflows in addition to writing software. The company says early versions of the model were even used to help build and test the final release itself. It is also faster than the previous version and available to paid ChatGPT users through the Codex app, command line tools, and development extensions.

Anthropic’s Claude Opus 4.6 focuses on a slightly different strength. The model improves long reasoning tasks, coding performance, and collaboration between multiple AI agents working together. It can maintain longer working memory and operate more reliably across large codebases, and it introduces features that allow several agents to coordinate on complex projects.

Anthropic has released a new version of its flagship AI model, Claude Opus 4.6, aimed at handling more complex tasks and appealing to a wider range of users. The update follows the launch of Opus 4.5 just a few months ago and focuses on expanding what the system can do inside its coding and workflow tools.

One of the biggest changes is the introduction of what the company calls agent teams. Instead of relying on a single system to handle everything, multiple AI agents can now work together on different parts of a task. They split work into smaller pieces, coordinate, and run jobs in parallel, which helps when dealing with large software projects or complicated workflows.

The model also comes with a much larger context window, allowing it to process far more information at once. This means it can understand bigger codebases and longer documents without losing track of details. Combined with stronger planning and task management abilities, the update is meant to make the system more reliable for long running or multi step work.

Opus 4.6 builds on the previous generation by improving accuracy and reducing the need for repeated rewrites, making it better suited for real production workflows rather than short experiments. The model is designed to support coding, enterprise tasks, and agent driven work where the system plans actions and coordinates tools on its own.

The release highlights how fast AI development tools are changing. Companies are no longer just improving how models answer questions. They are building systems that can divide work, collaborate internally, and handle larger responsibilities across software, research, and business operations.

Google has shared a new milestone for its AI assistant, Gemini. The company says the app now has more than 750 million people using it each month, based on results reported in its latest quarterly earnings update.

That number shows how quickly the service has grown since its launch. Just one quarter earlier, Gemini had about 650 million monthly users, meaning it added roughly 100 million in a short period.
This increase came after the rollout of Gemini 3, which Google described as an important step that drove stronger engagement and adoption among users.

The scale also reflects the broader shift toward AI tools in everyday consumer technology. Reaching this level of usage places Gemini among the largest AI assistants available today and highlights its role as a major competitor in the space, even though estimates suggest some rivals still have slightly larger audiences.

Google says activity inside the app has improved as well, with people spending more time interacting with it and using it more frequently after recent updates.
At the same time, the company has been working to lower operating costs by improving efficiency, which may help expand the service further across its products and platforms.

The milestone is less about a single feature and more about reach. It signals how deeply AI assistants are becoming part of daily digital routines. For Google, it confirms momentum. For the industry, it shows that adoption of conversational AI tools continues to grow at a rapid pace.

The AI writing partner, living in your Overleaf. No more LaTeX bugs or formatting hell. From tables to plots, Einsia handles the grunt work, reclaiming your time