Laplace bets AI red-teaming in simulation can make robots and self-driving cars safe

Builds automated red-teaming for robotics and physical AI: high-fidelity simulation of the real world plus AI agents that hunt for corner cases.

That the only way to make physical AI safe is to let AI test it — an agent probing interactive simulations to find failure points before deployment.

Howie Sun spent eight years in academia challenging autonomous-driving AI in simulation. In 2023 he and his then-PhD advisor Henry Liu published a paper that made the cover of Nature, and autonomous vehicle companies started calling. In 2025 the two founded Laplace Intelligence in Ann Arbor to turn the research into a product.

A year in, Laplace runs a post-trained version of Nvidia's Cosmos world model, grounded in real map data and live traffic information from providers like TomTom, where every actor reacts to every event — in one demo, an oncoming car swerved to dodge a swerving test agent. AI agents design corner cases from news and police collision reports. Customers include Japanese trucking maker Isuzu, Texas autonomous-trucking firm Bot Auto and Nvidia, alongside a free San Jose smart-city pilot with Nvidia and T-Mobile.

Generalized robots have no recorded history to mine for edge cases, so generative simulation testing becomes necessary rather than optional.

Safety evidence that customers, insurers and regulators can trust is the bottleneck for physical-AI deployment; testing is a wedge into every deployment.

Applied Intuition's $1.2 billion shows the category is real, but its playbook predates interactive world models — the startup is betting the paradigm shifts underneath it.

When industry starts calling a research lab, the market has already validated the idea; the founder's job is converting inbound project work into a product before a better-funded rival does.

As of 2026-10-13 Laplace was preparing its Startup Battlefield pitch at TechCrunch Disrupt in San Francisco (October 13-15), converting project-based contracts toward recurring revenue, and raising funds to scale against Applied Intuition.