SkildAI releases S1 robot model that learns unseen tasks from one video demo

deepakpathak · x · 2026-08-26

SkildAI has released S1, a foundation model designed to learn from a single example. It can perform 10-minute-long, unseen tasks based on one video prompt without any fine-tuning. Operating via in-context learning in real-time, this is hailed as a 'ChatGPT moment for robotics,' addressing the challenge of generalizing to new tasks.

Related event: Skild AI Unveils S1, a Robot Foundation Model That Learns New Tasks from a Single Video(9 posts)→

Original post →

More from Embodied

Embodied channel →