One video, then seven in a row
Google's new model learned cornhole from a raw video of a human playing, then wrote its own simulation code.

Google's Gemini 4 Argon learned cornhole by simply watching one raw video of a human playing. In an internal test, Argon wrote its own practice simulator from that video and worked out the body movements. Google DeepMind says Argon then taught Apptronik's Apollo humanoid to play.
Argon is the model that learned, and Apollo is the robot that threw. In a real-world trial, Apollo quickly sank seven consecutive shots, according to Google's blog. We have watched plenty of cornhole on TV and still miss the hole.
Sources: Google blog (Oct 13), Google DeepMind on X (Oct 13)
- Google blog (Oct 13): the test: "In an internal test, Argon watched a raw video of a human playing cornhole, mapped the spatial coordinates of the environment, and calculated the required whole-body joint movements. It then autonomously built the simulation code."
- Google blog (Oct 13): the result: "When transferred onto Apptronik's Apollo humanoid, the robot quickly sank seven consecutive shots in a real-world trial."
- Google DeepMind on X (Oct 13): DeepMind's own retell: "Gemini 4 Argon taught a robot to play cornhole just by watching a video ... and taught @Apptronik's Apollo humanoid to sink 7 throws in a row."
Current as of Oct 14, 2026. The image is an AI-generated illustration. Not affiliated with anyone.