Laplace Intelligence Uses AI to Stress-Test Robots and Self-Driving Cars Before Real-World Deployment

Laplace Intelligence is building automated red-teaming for robotics and physical AI - deliberately stress-testing systems to make them fail in simulation before they ever reach the real world.

"AI agents are moving from computers into the real world, and we don't really want physical AI to fail in the real world," CEO and co-founder Howie Sun told TechCrunch, pointing to the unpredictability of AI agents.

The Ann Arbor-based company, founded a year ago out of the University of Michigan by Sun and chief scientist Henry Liu, is competing in Startup Battlefield 200 at TechCrunch Disrupt in San Francisco this week.

At its core is a high-fidelity simulation reconstructed from the real world, paired with an AI agent that constantly probes for weaknesses. The environment is a post-trained version of Nvidia's Cosmos platform, grounded in real-world map data so environments are true to scale. Developers can change settings in real time with language prompts, and other actors in the simulation react to events rather than being driven through.

The startup's key selling point is AI agents that automatically design corner cases, drawing on news and police collision reports to challenge autonomous vehicles and other physical AI. Sun argues this matters most for generalized robots, where less data exists.

Laplace is generating revenue and a modest profit, mostly from autonomous driving customers including Texas-based Bot Auto and a major Japanese automaker. It is also working with Nvidia and T-Mobile on a free smart city pilot for San Jose, California, and is in early stages with warehouse humanoid robotics partners.

Its near-term goal is to raise funds and scale to compete with rivals, chief among them Applied Intuition. Sun believes using AI to test AI is a "new paradigm" - and "that will be the biggest moat."