{"id":660306,"date":"2026-10-08T11:08:52","date_gmt":"2026-10-08T09:08:52","guid":{"rendered":"https:\/\/preprod.aivancity.ai\/blog\/skild-ais-s1-what-video-based-learning-really-changes-for-robotics\/"},"modified":"2026-10-23T13:20:56","modified_gmt":"2026-10-23T11:20:56","slug":"skild-ais-s1-what-video-based-learning-really-changes-for-robotics","status":"publish","type":"post","link":"https:\/\/aivancity.ai\/blog\/en\/skild-ais-s1-what-video-based-learning-really-changes-for-robotics\/","title":{"rendered":"Skild AI\u2019s S1: What Video-Based Learning Really Changes for Robotics"},"content":{"rendered":"\n
\n

\n U<\/span>nveiled by\n Skild AI<\/strong> in\n August 2026<\/strong>, S1<\/strong> is a\n foundation model for robotics<\/strong> designed to\n perform a new task from a\n single video demonstration<\/strong>, without any additional\n task-specific training. The published trials\n show progress on\n manipulation sequences lasting up to ten minutes<\/strong>,\n but they still rely on\n internal evaluations<\/strong>. The stakes therefore go beyond\n the technical demonstration: the question is whether this\n in-context learning<\/strong> can become\n reliable, safe and\n controllable enough for professional environments.<\/strong>\n <\/p>\n<\/div>\n\n

\n\n
\n 01\n <\/div>\n\n

\nWhat Just Happened\n <\/h2>\n\n<\/div>\n\n

Skild AI unveiled S1 in August 2026 as its flagship foundation model for robotic manipulation. The system receives a video showing a task, then attempts to reproduce the observed goal with a robot, without changing its weights and without any task-specific post-training phase. On September 10, NVIDIA detailed the infrastructure used to train, simulate and deploy the model, including Cosmos, Omniverse, Isaac Sim and Isaac Lab.[1][<\/a>2]<\/a><\/p>\n\n

The promise concerns new, long-horizon tasks. Skild AI shows S1 repotting a plant, brewing filter coffee, making cr\u00eapes or assembling a kit. Some sequences last up to ten minutes and involve several successive movements. For the repotting task, the company says it went from recording the demonstration to the start of autonomous execution in eleven minutes.[1]<\/a>[2]<\/a><\/p>\n\n

This timeline needs clarifying: S1 was not launched on September 10. The initial announcement dates back to August 2026, while NVIDIA\u2019s September 10 publication presents the technical collaboration and the use of its computing and simulation platforms.<\/p>\n\n

\n\n
\n 02\n <\/div>\n\n

\nWhat Is Actually New\n <\/h2>\n\n<\/div>\n\n

The idea of teaching a robot a movement by demonstration is not new. Programming by demonstration, imitation and teleoperation have been studied for decades. What sets S1 apart is its ambition to bring to robotics a mechanism that has become central to large language models: in-context learning.<\/p>\n\n

In this framework, the video is not used to retrain the model. It is placed in the model\u2019s context at the time of use. S1 must extract the intent from it, recognize the relevant objects, track the task\u2019s progress and translate the demonstration into actions compatible with the robot and the scene at hand. The company states that the same set of weights produces all the examples shown.[1]<\/a><\/p>\n\n

The claimed novelty also concerns the length and novelty of the tasks. Skild AI says that several scenarios did not appear in the pre-training data and that they combined basic movements in new sequences. If this capability is confirmed outside internal trials, it could reduce the cost of data collection and specialization for each new operation.<\/p>\n\n