Skild AI

Skild AI is asking a bold question: can one AI brain control any robot for any task? That's a fundamentally different approach from most robotics companies, which specialize. The idea of learning from human videos is the key differentiator, because it sidesteps the expensive, robot-specific data collection that stalls most generalist efforts. If Skild pulls this off, the payoff is enormous. But that's a big 'if'. Here's the catch: there are no public demos, no technical specs, no deployment evidence. The website is a vision statement, not a product showcase. For most buyers, that's a dealbreaker. You cannot evaluate a tool you cannot see. If you need robots working in your factory or warehouse this quarter, Skild is not the answer. The company's three pillars—security/inspection, mobile manipulation, and autonomous packing—are smart because they target real-world value. But they are just descriptions, not shipped solutions. The API abstraction for mobile manipulation is a promising idea, but without documentation or examples, it's vaporware from a user's perspective. Where Skild could fit is as a research partner. If you're an innovation lab or a university exploring generalist physical AI, Skild might offer valuable collaboration. The venture backing suggests they have runway to iterate. But even then, you'll be betting on the team's ability to execute, not on a proven product. Compared to Boston Dynamics, which has public demonstrations of highly capable legged robots, Skild is a leap of faith. Covariant has real-world pick-and-place deployments. Skild has none. For most organizations, that asymmetry matters.