Figure 02 Humanoid Robot Demos Shocking 16 DoF AI Automation (NEW OLI ROBOT)

Figa and Limex Dynamics are developing advanced home robots that can perform tasks like laundry, unpacking groceries, and sorting packages. These robots use a system called Helix to control their movements and understand language. Meanwhile, Tiankong 2.0, a robot with all-terrain capabilities, uses artificial intelligence to perform complex tasks and adapt to changes in its environment. Google and Anthropic are also making strides in AI, with Google updating its Gemini application and Anthropic introducing a new feature that allows users to create applications and content without coding.

Are home robots about to become mainstream? Figa is pushing the boundaries of home robot automation once again, now demonstrating its humanoid’s ability to autonomously handle laundry, adding to its existing repertoire of abilities which include grocery unpacking, sorting packages, and preparing coffee. In fact, the robot’s ability to perform these real-world tasks is powered by what Figa calls Helix, a generalist vision-language action model designed to unify perception, language understanding, and continuous control across the robot’s upper body. But Helix introduces several new advancements to make it all possible. For example, this is the first time that a VLA model has enabled full upper body control, orchestrating synchronized movements of the wrists, torso, head, and even individual fingers.

This allows the robot, standing 5 feet 6 inches tall and weighing 70 kilograms, to manipulate a wide variety of small household objects and interact with appliances such as laundry machines, all while carrying up to 20 kilograms and moving at 1.2 meters per second. Plus, with a runtime of five hours on a single charge and an all-electric system, Figa’s robot is designed for extended practical use in home or commercial settings. Meanwhile in China, Limex Dynamics just announced its newest Ollie humanoid robot, showcasing a suite of advanced features for the real world. To demonstrate its abilities at the company’s headquarters, the Ollie humanoid was observed performing whole-body mobile manipulation, including precise material handling and sorting, highlighting its capabilities in dynamic work environments.

But another notable aspect is the Ollie robot’s use of self-developed joint modules, which are engineered to provide stable torque output. This contributes to smooth and controlled movements, even during physically demanding tasks. On top of this, the robot’s hands are highly adaptable, with three available out fittings, standard hand models, robust industrial grippers, and fully dexterous anthropomorphic hands, allowing Ollie to manipulate a wide variety of objects, but the robot’s lifting capacity is still unknown. Furthermore, Ollie employs natural language processing, allowing it to interact and communicate effectively with humans in its environment. This is complemented by its full-body disturbance rejection system, which helps the robot maintain balance, even when subjected to unexpected pushes or kicks.

Plus, Ollie seamlessly switches between multiple gate patterns, enabling it to adapt its walking style for different surfaces and obstacles. For now, though, technical specifications for the Ollie robot have not yet been released by Limex Dynamics, so tell us in the comments how much you think it’ll cost. But competition is heating up in China, as another demo of the new Tiankong 2.0 just dropped showcasing its all-terrain capabilities and deep integration of artificial intelligence. As for specs, the robot weighs over 100 kilograms, with its exact weight still being unspecified, and features a total of 35 degrees of freedom, enabling highly dynamic, human-like movement across uneven terrain.

On top of this, it boasts an all-day battery life for uninterrupted performance, making it a reliable solution for missions that demand extended operation without human intervention. Furthermore, Tiankong 2.0 features an advanced embodied intelligent system powered by a multimodal VLA large model. This endows the robot with deep visual understanding, spatial awareness, and the ability to interpret and execute complex task instructions. One of its standout capabilities is precise task decomposition with dual-arm collaboration, allowing it to break down multifaceted assignments, select the proper tools, and perform coordinated bi-manual tasks with high accuracy.

Plus, its real-time error correction and adaptive re-planning allow the robot to instantly adjust to sudden changes in its environment or mission requirements. Its comprehensive suite of sensors and AI-driven perception systems ensures safe navigation around people and obstacles, making the IT a promising new standard for robust, intelligent humanoid robots. And in the latest and greatest in generative AI, Google has introduced a notable update to its Gemini application programming interface, now offering both the VO3 fast model and a new image-to-video feature. The VO3 fast version is engineered to prioritize speed and cost-effectiveness, specifically tailored for developers needing to iterate quickly or produce content at scale, such as for programmatic advertising or rapid A-B testing scenarios.

According to Google, VO3 fast still maintains a high-quality standard, despite its efficiency-focused design. Both VO3 and VO3 fast provide the same technical capabilities. Users can submit text or image prompts, resulting in video clips at 720 pixel resolution and 24 frames per second. Each request generates a video up to eight seconds long, with a maximum of one video per request. Plus, both models support up to 1024 tokens per text input and include native audio generation. And Google’s new image-to-video feature is available in both versions, allowing users to combine a single image with a text prompt to generate dynamic videos complete with audio.

This addition is designed to enable greater creative control, letting creators maintain stylistic consistency and direct movement, sound, and narrative flow through their prompts. Furthermore, the pricing structure reflects VO3 fast’s efficiency, at 40 cents per second of video with audio, compared to 75 cents per second of video without audio second for the standard VO3. For example, an eight-second clip costs $3.20 with VO3 fast, while the same video costs $6.40 with the original VO3, marking a significant difference for large-scale or budget-sensitive projects. And for vibe coders, Anthropic has taken a major step forward with Claude artifacts, introducing the ability for anyone to transform ideas into shareable applications, tools, and content simply by describing what they want.

Users can build interactive tools, data visualizations, and customized experiences without needing to write code. Plus, Claude offers a dedicated window for these artifacts, keeping substantial standalone content separate from the main conversation for easier focus and organization. Users can also revisit, modify, and share their creations, making collaboration and iteration straightforward. Overall, the update means complex projects are now more accessible to a wider range of users. By streamlining the process from concept to final product, Anthropic is making it easier than ever for anyone to bring new ideas to life, all within Claude’s AI-powered platform.

And finally, Manus just announced wide research. This new feature allows users to tackle large-scale research tasks by enabling high-volume, parallel information gathering and processing. By providing each user with a dedicated cloud-based virtual machine, anyone can now orchestrate complex workloads simply by conversing with an agent. What’s more is that each Manus session harnesses a Turing-complete environment, enabling everything from crafting tailored rental presentations to securely running open-source projects in a sandboxed space. Furthermore, wide research moves beyond traditional multi-agent systems, as every sub-agent is a fully capable Manus instance, allowing for flexible, domain-agnostic collaboration.

Anyways, like and subscribe for more AI news, and check out this video here. [tr:trw].