This Week in AI: Largest Humanoid Factory

Apple AirPods to Add Live Translation Feature

Apple reportedly plans to add live language translation capabilities to future AirPods wireless earbuds. The feature would compete with similar functionality already available in Google's earbuds.

Google Launches Gemma 3 Model

Google DeepMind releases Gemma 3, a compact AI model that outperforms much larger systems while running on single GPUs. The model supports 140 languages and processes multiple media types.

Chinese Team Creates First Stable 2D Metal

Chinese researchers produce the first stable, large-area single-atom thick metal layer. The breakthrough could revolutionize transistors, displays, sensors, and catalysis technology.

Manus: The Action-Taking AI Assistant

Manus introduces a general AI agent that both analyzes and completes complex tasks independently. The system achieves record scores on GAIA benchmarks.

Create 3D Art Through Natural Language

A new MCP enables Claude to control Blender software through natural language commands. Users can create detailed 3D scenes by describing what they want.

OWL: Free Alternative to Premium AI Agents

OWL, an open-source alternative to Manus AI, gains popularity on GitHub. The framework supports multiple AI models including Claude, DeepSeek, and local LLMs.

AI Robot Demonstrates Cycling Skills

A humanoid robot demonstrates the ability to ride a bicycle, showcasing remarkable balance and coordination. This achievement represents a significant breakthrough in robotics and physical AI.

Three-Minute Call Gets Startup $100K from AI

An AI called No Cap autonomously invested $100,000 in a startup after just a three-minute call. The female-presenting system signs agreements and provides business connections independently.

Sesame Labs Releases Open Speech Model

Sesame Labs releases CSM, an open-source speech model trained on one million hours of data. The Apache-licensed system offers emotional intelligence and voice cloning capabilities.

Vogent Agents Fix Their Own Mistakes

Vogent launches beta for self-learning voice agents that improve automatically from call recordings. The system fixes its own mistakes without requiring manual prompt engineering.