AIToday
RoboticsThe Robot ReportPublished: Sep 1, 2026, 10:00 JST2 min read

Skild AI unveils S1 robot foundation model

Skild AI unveils S1 robot foundation model

Key takeaway

  • Skild AI has launched S1, a robot foundation model that learns tasks from a single video.

  • It combines all four data types for pre-training.

  • The company aims to scale deployment with its Fetch Robotics acquisition.

3 Key Points

  1. What happened

    Skild AI has introduced S1, its flagship robot foundation model, which lets robots learn new, complex tasks by watching a single video. The company claims S1 can handle tasks lasting up to 10 minutes, like making coffee or cooking pancakes.

  2. Why it matters

    Traditionally, robots must be post-trained for each new task, which is slow and limits scaling. Skild AI combines all four data types — teleoperation, human videos, simulation, and data-capture gloves — to build a general-purpose robot brain, aiming to move beyond single-task robots.

  3. What to watch

    Skild AI has raised nearly $1.7 billion since 2023 and recently acquired Fetch Robotics to boost deployment. The company says S1 is already helping in production and plans to show more in the coming weeks, though it admits robotics is not yet at a "ChatGPT moment" for home use.

Ask the AI about this article →

Context & Analysis

Skild AI's S1 model represents a shift in robotics, moving away from task-specific fine-tuning toward a more general approach. By leveraging all four data types, the company addresses the limitations of each, such as the diversity of human videos and the directness of teleoperation data. This mirrors the transition in language models from fine-tuning to prompt-based learning, though Pathak acknowledges robotics still has a way to go before achieving a 'ChatGPT moment.'

The company's focus on deployment, highlighted by the Fetch Robotics acquisition, indicates a strategic push to bring S1 to practical use. Pathak emphasizes that deployment is a high priority, and S1 is already helping in production to acquire more customers. However, the model's performance on humanoids is still in early stages, with results from an earlier version showing adaptability but not yet scaled.

Skild AI's ambition to create a general-purpose robot brain is backed by significant funding and a clear vision. The coming weeks will reveal more about S1's production capabilities, offering a glimpse into whether robotics is on the brink of a transformative leap similar to what ChatGPT did for language AI.

FAQ

What makes S1 different from other robot AI models?
S1 uses in-context learning, so robots can follow a new task from a video prompt without post-training. It also combines all four data types to pre-train the model, unlike most companies that focus on one.
Is S1 ready for home use?
Not yet. Skild AI's CEO says it's not completely ready to be rolled out to people's homes, but it is a first sign of what might come.
What tasks can S1 handle?
S1 focuses on longer tasks that can take up to 10 minutes, such as repotting a plant, making a cup of coffee, or cooking pancakes.
The Robot ReportRead Original Article

Get the latest Robotics news every morning

For example, today's edition would include:

  • BHP's AI cuts downtime, boosts iron ore outputTop Companies AI · 10h ago
  • John Deere invests $10M in AI for high-value cropsTop Companies AI · 10h ago
  • NVIDIA expands Cosmos Coalition to Japan with 22 companiesTop Companies AI · 10h ago

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleNvidia invests ~¥560B in Taiwan's MediaTek