Skild AI Unveils S1 Robot Foundation Model to Learn Tasks From a Single Video

Industrial automation frequently breaks when assembly lines shift and layouts change, requiring tedious reprogramming for each new movement. Skild AI's S1 foundation model addresses this friction by executing unseen, long-horizon tasks from a single video demonstration using in-context learning—without fine-tuning or updating model weights. Developed on NVIDIA's AI and simulation infrastructure, including Isaac Lab and Cosmos, the system shifts robot adaptability from per-task engineering toward zero-shot visual prompting in physical environments.