Google DeepMind Adds a Robotics Plugin to Antigravity So Gemini 4 Hydrogen Can Teach Robot Arms Tasks in Plain Conversation
Teaching robots new tasks just got a lot faster. 馃 With Gemini 4 Hydrogen and the new robotics plugin for Google @Antigravity, you can connect a supported arm, practice in @MuJoCo simulation, and run tasks in the real world using plain conversation.
Context
Connecting a robot arm, practicing in simulation and running the task on real hardware through a coding tool lowers the skill needed to program robots.
Antigravity is Google's agent based coding environment, and Gemini 4 Hydrogen is one of Google's current Gemini models. MuJoCo is a physics simulator maintained by Google DeepMind and widely used in robotics research. Google DeepMind announced the robotics plugin on Nov 17.
Deep dive
Google DeepMind posted that teaching robots new tasks just got a lot faster. According to the post, with Gemini 4 Hydrogen and a new robotics plugin for Google Antigravity, a user can connect a supported robot arm, practice in MuJoCo simulation, and then run tasks in the real world using plain conversation.
The attached video opens with a person plugging a small robot arm into a laptop, with an on screen label reading Hydrogen and Plug in YAM arm. A depth camera on a stand points at the desk. Only the opening frame was reviewed here, so the rest of the demonstration is not described.
The post does not list which arms are supported, whether the plugin is available to all Antigravity users, or whether it costs extra. It also does not give success rates or say how well skills practiced in simulation carry over to the physical arm, which is a common difficulty in robotics. Several replies to the post asked about that gap between simulation and the real world.
The launch fits a pattern in which coding agents are being extended beyond software. On Nov 16, the Gemini app showed Hydrogen and Antigravity turning a sketch into a font, and on Nov 17 Genesis announced a $500 million round for general purpose robots, also covered in this deck.
What to expect next
Watch for documentation that lists supported arms and any hardware partners, and for whether the plugin reaches all Antigravity users or stays limited. Early user videos will show how reliably tasks transfer from MuJoCo simulation to real hardware. Competing coding agents may add similar hardware plugins if this one gets traction with hobbyists and research labs. Google may also say which other Gemini models can drive the plugin.
