Visions of AI: GPT-3 Moment for Physical AI

Skild AI released S1, an in-context learning model for robotics that the company claims is a foundation model capable of learning 10-minute tasks from a single video prompt without fine-tuning, and the startup has surpassed a $100 million revenue run rate in 2026, according to Dealroom. The release follows a late-2026 spike in Physical AI and embodied AI funding rounds and a wave of Chinese robotics IPOs, including Unitree's pricing at a $9 billion valuation. AI Supremacy's analysis argues S1 could qualify as a GPT-3 moment for robotics learning in its 2027 iterations, comparable to OpenAI's 175-billion-parameter GPT-3 paper published May 20, 2020, roughly 3.5 years before ChatGPT.

Visions of AI: GPT-3 Moment for Physical AI AI Supremacy https://www.ai-supremacy.com Has embodied AI and robotics hit a milestone in 2026? Good Evening, Visions of AI is a new feature format I’m experimenting with that will amount to a short profile on an AI or emerging tech startup. The cadence of this style of article is unknown as of yet, but there are a lot of fascinating startups I want to discuss and share about. This is designed to be light evening reading to go out at a time-slot of 8 pm EST. In late 2026 there’s been a very sudden spike in massive Physical AI and embodied AI funding rounds, especially in the software around the robotic brain, and humanoid robotics https://www.machinebrief.com/category/robotics , a form factor of bipedal human-like robots. While it remains to be seen how significant or enduring this trend is, it follows a flurry of Chinese robotics https://www.cnbc.com/2026/08/06/chinese-humanoid-robot-maker-unitree-prices-ipo-at-9-billion-valuation.html companies going public. China is widely seen as the robotics leader of the world, especially in current sales, hardware and manufacturing. About three weeks ago a Physical AI startup https://www.machinebrief.com/category/startups called Skild AI released their S1 https://www.skild.ai/blogs/s1 , an in-context learning https://www.machinebrief.com/glossary/in-context-learning model for robotics. Skild AI has hit over $100 million https://dealroom.co/news/150181-skild-ai-hits-100m-revenue-run-rate-10-months-after-first-deployment/ 1 footnote-1 in revenue https://dealroom.co/news/150181-skild-ai-hits-100m-revenue-run-rate-10-months-after-first-deployment/ already in 2026. The company claims that S1 new is a foundation model https://www.machinebrief.com/glossary/foundation-model that learns from one example one-shot learning . They said that it can be taught 10-minute long tasks that it has never seen before, from one video prompt without any fine-tuning https://www.machinebrief.com/glossary/fine-tuning . Embodied AI with a Real Revenue Ramp Skild After doing some digging while it’s a marketing idea being pushed by the industry , I do believe S1 could qualify as the GPT-3 moment for robotics learning, at least in its 2027 iterations. In the history of AI, that would correspond to around 2020. Generative Pre-trained Transformer https://www.machinebrief.com/glossary/transformer 3 GPT-3 was a landmark 175-billion parameter https://www.machinebrief.com/glossary/parameter LLM developed by OpenAI where the research paper introducing it was published on May 20th, 2020, about 3.5 years before the ChatGPT https://www.machinebrief.com/compare/chatgpt-vs-claude moment itself. This is tremendously exciting for robotics and Physical AI enthusiasts. Get AI news in your inbox Daily digest of what matters in AI. Key Terms Explained Fine-Tuning https://www.machinebrief.com/glossary/fine-tuning The process of taking a pre-trained model and continuing to train it on a smaller, specific dataset to adapt it for a particular task or domain. Foundation Model https://www.machinebrief.com/glossary/foundation-model A large AI model trained on broad data that can be adapted for many different tasks. GPT https://www.machinebrief.com/glossary/gpt Generative Pre-trained Transformer. In-Context Learning https://www.machinebrief.com/glossary/in-context-learning A model's ability to learn new tasks simply from examples provided in the prompt, without any weight updates.