This simulation startup wants to be the Cursor for physical AI

Antioch, a New York-based simulation startup, has secured $8.5 million in seed funding at a $60 million valuation to tackle a core bottleneck in robotics: the lack of scalable, realistic training data. Led by founders with backgrounds at Meta Reality Labs, Google DeepMind, and a previous successful exit, the company is building a platform designed to be the "Cursor for physical AI." It allows developers to create high-fidelity digital replicas of robots and their environments, connecting them to simulated sensors to generate the data needed for training and testing AI models without building costly physical mock-ups.

The company's mission is to close the notorious "sim-to-real gap," ensuring virtual physics match reality so robots trained in simulation can operate reliably in the real world. Antioch builds on simulation models from partners like Nvidia and World Labs, creating domain-specific libraries to make them accessible. By working with multiple robotics customers, Antioch gains diverse context to refine its simulations, a depth no single company could achieve alone. This approach aims to democratize advanced simulation, enabling smaller companies without the capital for massive physical testing to accelerate development, much like AI coding tools have done for software engineers.