📰 Key Takeaways

TechCrunch Disrupt 2026 runs October 13-15 at Moscone West in San Francisco, and for the first time it’s splitting AI content into two stages: the existing AI Stage keeps its usual lineup, while a new “Real World AI Stage” digs into where the digital and physical worlds collide — how autonomous hardware beyond self-driving cars is showing up in public spaces, on battlefields, in homes, and even helping bring extinct species back. Confirmed speakers include reps from Shield AI, Colossal Biosciences, FieldAI, and Foxglove, with three headline sessions planned so far. First up, Shield AI CTO Nate Michael on “building AI systems when failure isn’t an option” — how hard-tech founders in self-driving, defense tech, and industrial systems decide when a system is safe to deploy, build a safety culture, and get through regulatory review. Second, a fireside chat with Colossal Biosciences CEO Ben Lamm on how his company uses AI to bring back extinct species, AI’s role in modern biology, and whether “engineering nature” is a conservation breakthrough or a distraction that drains resources from real conservation work. Third, FieldAI CEO Dr. Ali Agha, Medra CEO Michelle Lee, and Eclipse Ventures partner Aidan Madigan-Curtis will discuss how AI operates in edge environments cut off from cloud connectivity — covering latency and connectivity constraints in defense, space, and industrial settings. Buy tickets through the official link before 11:59pm Pacific on August 7 for an extra $100 off.

💬 JudyAI Lab Take

This year’s TechCrunch Disrupt is splitting AI into two stages for the first time, and the new “Real World AI Stage” shifts the focus from pure software to AI that actually acts in the physical world — everything from defense hardware to endangered species revival is on the table. The split itself is a signal worth noting.

The speaker lineup points to a bigger shift: the hard problem in AI development is moving from “is the model accurate” to “can the system safely go live.” Shield AI is talking about building a safety culture and getting through regulatory review when failure is extremely costly, while FieldAI has to handle how latency and disconnection affect decision-making in edge scenarios with no cloud access. These aren’t problems you solve by scaling up the model — they’re engineering and governance tradeoffs, and anyone building agents or automation systems should be thinking about them early.

Figuring out your system’s fallback plan for when it loses connection or makes a bad call is worth more right now than rushing to train your next model version.

📅 Source Info

🔗 Further Reading