TechCrunch Disrupt 2026 adds Real World AI Stage to split AI platform

The conference opens on 13 October 2026 in Moscone West, San Francisco. Organizers will launch a “Real World AI Stage.” The separate stage will focus on the intersection of digital and physical realms. It targets autonomous hardware moving from self-driving cars to public spaces, battlefields, private homes, and de-extinction projects. Speakers include representatives from Nvidia, Shield AI, Colossal Biosciences, FieldAI and Foxglove, with additional participants yet to be announced.

Les Karpas, head of Physical AI at Nvidia, will lead a session on why general-purpose robotics still lags behind breakthroughs in natural language. He will argue that large language models benefited from the internet as a massive training source, and autonomous vehicles have millions of logged driving hours. Robots have none. This data gap is defined as the single biggest obstacle. A new wave of startups is trying to close it with data pipelines, simulation environments and purpose-built foundation models.

Nate Michael, senior vice president of technology at Shield AI, will discuss the stark difference between a chatbot error and a failure in a physical system—grounded aircraft, vehicle crash or botched military mission. The session will examine how to know a system is ready for safe deployment. Topics include building a safety culture, rigorous validation and testing, navigating regulatory hurdles and earning customer trust when the stakes are high. The talk targets hard-tech founders who cannot afford to “move fast and break things.”

Ben Lamm, CEO of Colossal Biosciences, will appear for a fireside chat on turning de-extinction from science-fiction into a billion-dollar business. The discussion will cover technologies that enable species restoration, the role of AI in modern biology, and the growing public debate: Is engineering nature a breakthrough for conservation or an expensive distraction from protecting existing life? Lamm will need to defend his business model against scientific and ethical criticism.

A session chaired by Dr. Ali Agha, founder and CEO of FieldAI, together with Michelle Lee, founder and CEO of Medra, and Aidan Madigan-Curtis, partner at Eclipse Ventures, will address AI deployments in disconnected or bandwidth-limited settings such as defense, space and heavy industry. Participants will share architectural principles, design decisions and trade-offs that let systems operate when latency is critical, bandwidth is constrained and failure is not an option. The playbook differs fundamentally from public-cloud approaches.

The closing session will confront the familiar gap for hardware builders: a lab-working prototype is not a product, and a shipped product does not automatically scale. Speakers will focus on bottlenecks that appear when moving an autonomous system from low-volume to mass production while preserving reliability, cost and delivery timelines. The challenge causes many young hard-tech companies to stumble.