Laplace Intelligence red-teams robots and self-driving cars before they hit the real world

Laplace Intelligence, an Ann Arbor, Michigan startup spun out of University of Michigan research, is building automated red-teaming for robotics and physical AI, and it will pitch in the Startup Battlefield 200 competition at TechCrunch Disrupt in San Francisco on Tuesday afternoon, October 13. The event runs October 13-15. The company's co-founders are CEO Howie Sun and chief scientist Henry Liu, Sun's former PhD adviser; Sun says they founded the company a year ago. In 2023 the pair published research on challenging autonomous driving AI in simulation that landed on the cover of Nature, which Sun says prompted inbound interest from autonomous vehicle companies.
Red-teaming means deliberately attacking a system to surface failures. Laplace applies that idea to machines that operate in the physical world: an AI agent probes a simulated robot or vehicle for weaknesses, and problems are meant to be found and fixed before real-world deployment. Sun frames the company's longer-term goal as assembling a shared base of evidence that customers, insurers and regulators can rely on. The name references Laplace's demon, Pierre-Simon Laplace's 1814 thought experiment about predicting the future from complete knowledge of the present; Sun describes the company's work as using simulation to predict the real world's future.
The core product is a high-fidelity simulation reconstructed from the real world, combined with an agent that continuously searches for corner cases. The simulation builds on a post-trained version of Nvidia's Cosmos world simulation platform. Laplace blends real-world inputs such as map data and Street View imagery to regenerate environments at true scale, from an Arizona highway to a Paris pedestrian crossing, and developers can adjust conditions with natural-language prompts, for example adding afternoon traffic. Sun says the system pulls live traffic and congestion information from navigation providers such as TomTom. Other actors in the scene are simulated behaviorally as well as visually: in a demo, when Sun steered the test vehicle into oncoming traffic, the oncoming car swerved to avoid it, an outcome Sun contrasts with generative world models where vehicles sometimes pass through each other. He describes it as an interactive environment where every participant reacts to everything else.
The differentiating piece, in the company's telling, is the AI agents that automatically design tricky, rare scenarios, reducing reliance on humans. Sun says the agents can draw on sources such as news and police collision reports to challenge autonomous vehicles and other physical agents. Today, autonomous vehicle developers typically build edge-case tests from recorded footage turned into simulation, or rely on people to invent scenarios; several AV companies have begun using AI to generate edge cases, with self-driving truck developer Waabi among them. Sun argues the need will be sharper for general-purpose robots, where there is little data and nobody knows what the real-world edge cases will be.
Commercially, Laplace is generating revenue and, despite its academic roots, posting a modest profit. Most revenue comes from autonomous driving: customers include Texas-based autonomous trucking firm Bot Auto and a major Japanese automaker that the company has not named. Laplace is also working with Nvidia and T-Mobile on a free smart city pilot for San Jose, California, aimed at high-level traffic operations, and it is in early work with partners building warehouse humanoid robots. Most contracts are project-based, and the company is working to convert them into recurring revenue; a near-term goal, which it hopes the Disrupt competition will help with, is raising money to scale against rivals.
The most prominent competitor is Applied Intuition, a late-stage startup that has raised $1.2 billion to date at a $15 billion post-money valuation. Sun is not deterred, arguing that testing AI with AI is a new paradigm that turns physical AI testing into a dynamic, interactive process, and that this will be the company's biggest moat.
Read as a product: Laplace is selling a simulation and scenario-generation layer, not a robot or a car, so its value depends on how faithfully its environments and agent behaviors track reality and on whether customers accept agent-generated corner cases as credible evidence. The pitch to insurers and regulators suggests a possible second business in certification-style validation, though that depends on acceptance the startup has not yet secured. Turning project work into recurring subscriptions is a familiar challenge for simulation vendors, and the valuation gap with Applied Intuition shows how much capital separates early-stage entrants from incumbents in this market.
Why it matters: Teams building robots and autonomous vehicles get another option for automated scenario generation aimed at rare, dangerous situations, which could shorten test cycles and reduce manual scenario authoring (an inference from the company's stated approach). Because Laplace is betting on general-purpose robotics as the next frontier, warehouse and humanoid robot developers may soon see simulation testing marketed directly to them. For product leaders, the open question is whether agent-designed scenarios become trusted evidence for safety cases and insurers, which would change how autonomy programs document readiness.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
University of Michigan spinout Laplace Intelligence uses AI agents and interactive simulation to stress-test robots and self-driving vehicles before deployment. The profitable startup is competing in Startup Battlefield at TechCrunch Disrupt, October 13-15 in San Francisco.