Updated
Updated · TechCrunch · Aug 26
Foxglove Launches Nvidia Cosmos Search Tool for 1,500-Strong Physical AI Market
Updated
Updated · TechCrunch · Aug 26

Foxglove Launches Nvidia Cosmos Search Tool for 1,500-Strong Physical AI Market

3 articles · Updated · TechCrunch · Aug 26

Summary

  • Foxglove unveiled a new product built on Nvidia’s Cosmos that lets robotics engineers search visual and lidar data with natural-language queries to speed debugging, evaluations and simulation workflows.
  • The launch targets a core bottleneck in physical AI: scarce high-quality training data and robots that still struggle to deliver reliable commercial performance despite improving hardware.
  • 1,500 developers attended Foxglove’s Actuate conference last week—triple its 2023 size—highlighting investor and builder enthusiasm even as the sector debates whether general-purpose humanoids are still years away.
  • That tension has sharpened after Unitree’s $66 billion IPO valuation was followed by a near-halving this week, underscoring doubts that current robot AI can yet turn technical progress into value-creating work.

Insights

If data is the real bottleneck for physical AI, why are autonomous vehicle giants pivoting to humanoid robots now?
As startups rush to solve the robotics data crisis, could biological learning models render their massive simulation engines obsolete?
Will the promised breakthrough for robotics be driven by massive data centers, or by a simple machine cleaning your living room?