Skild AI Unveils S1: A Foundation Model That Learns Robot Tasks From a Single Video

Skild AI has unveiled S1, its flagship robot foundation model, and the headline claim is one that should get the attention of every engineer working on flexible automation: robots running S1 can learn new, complex tasks from watching a single video, thanks to in-context learning.

No Fine-Tuning Required

Typically, when AI models on robots face new tasks, they need to be post-trained to handle them — a lengthy process that has held robotics back from scaling. S1 takes a different approach.

"You just add a video of a human doing something in the prompt, also called the context of the model, and it can just follow it on the robot," said Deepak Pathak, Skild AI co-founder and CEO.

Pathak was emphatic that the demonstrated tasks are not trivial:

"The tasks we are showing are extremely complex and long horizon. They are not three-second, four-second tasks, not those tiny, simple tasks."

Four Data Sources, No Golden Path

According to Pathak, there are four kinds of robot training data, and Skild AI makes use of all of them: teleoperation data, human videos, simulation, and data-capture gloves. Each has tradeoffs — teleoperation is high quality but slow to capture, human video is abundant but hard to apply directly to robots, simulation is scalable but has a real-world gap, and gloves sit somewhere in between.

"If you look at any company out there, they are primarily focusing on one of these sources. But, if we think from first principles, we realize there is no golden path," Pathak said. "We have to use all of them because the pros of one source compensates for the downside of another source."

Omni-Bodied and Task-Agnostic

Skild AI is not focusing on any one specific task or industry. Instead, it targets longer tasks — up to 10 minutes — like repotting a plant or making a cup of coffee. The brain is also "omni-bodied": it can work on anything from a quadruped to a humanoid or a static arm, with more humanoid-focused work planned for the future.

Deployment Is the Next Frontier

Is robotics having its ChatGPT moment? Pathak doesn't think so — not quite yet.

"Now, the question is: Is it completely ready to be rolled out to people's homes? Not quite. But this is the first sign of what we believe might come."

The company's acquisition of Fetch Robotics — whose assets came from Zebra Automation — was about getting the right talent to deploy and scale. Skild AI says it is focused on both scaling frontier research and deployment, with production evidence coming soon.

"You will see in the coming weeks, we'll show how S1 is already helping in production," Pathak said. "It's actually helping us move faster to acquire more customers."

For manufacturing professionals, that's the signal to watch: a generalist robot brain is only as valuable as its deployment track record.