Skild AI has unveiled S1, which it describes as a robotics foundation model, according to a Techmeme item linking to Skild AI’s announcement. The company says the model is designed to learn tasks that were not seen during pretraining. The core claim is narrow but notable: Skild AI says S1 can use a single video demonstration, or “one video prompt,” to learn a new robotics task without fine-tuning or other post-training. The summary also highlights “10-minute horizons,” though it does not provide enough detail to explain the benchmark or operating conditions behind that phrase. For now, this should be read as a company announcement rather than an independently validated model result. The provided source material does not include outside testing or benchmark comparisons. Who benefits: Robotics teams looking for faster task adaptation could benefit if S1 performs as described. Skild AI benefits from positioning S1 as a foundation-model approach to general-purpose robotics. Who's exposed: The provided source material does not provide enough evidence to identify specific companies or products that are directly exposed. The broader risk would be to robotics approaches that require more task-specific training, but that remains unproven from the available material.