Skild AI Launches S1 Robot Foundation Model With Single-Video Learning

Skild AI has launched S1, a robot foundation model that uses in-context learning to execute unfamiliar multi-step tasks from a single video demonstration without retraining or weight updates. The model adapts behavior by changing only the video prompt, a property described as "prompt steerability," which determines whether a new robot task requires new training. In tests, S1 succeeded on roughly 66% of steps for new multistep tasks compared to roughly 9% for a comparable AI system, and one video demonstration was estimated equivalent to roughly 380 hands-on training examples. Skild, NVIDIA, and Foxconn are deploying the model on dual-arm manipulators for high-precision assembly of NVIDIA Blackwell systems. Skild and NVIDIA are also jointly developing GPU-accelerated simulation solvers for robot manipulation to be released via Newton. Skild AI reached $100 million annual revenue run rate within 10 months of its first commercial deployment and has over 60 deployment partnerships.