Real Alpha

Latent Space · 4 min instead of 1 h 21 m · Oct 13, 2026

Multi-site rollouts, not pilots, prove a robot creates value

Jason Ma

Co-founder & chief scientist, Dyna Robotics

First author of Eureka, in which GPT-4 wrote robot reward functions that beat expert ones; previously at Google DeepMind and NVIDIA.

  • Shawn Wang (swyx)Host · Co-founder, AI Engineer

    Wrote “The Rise of the AI Engineer” and founded Smol AI, maker of AI News.

  • Alessio FanelliHost · Founder, Kernel Labs

    Partner at Decibel, backing technical founders who build developer tools.

About this segment

Jason Ma describes Dyna robots folding thousands of napkins daily across Din Tai Fung locations, which led to a larger contract. He argues pilots are easy; customers expanding to many sites is the real test.

From “Robots That Predict the Future: Dyna-2 & Physical AI — Jason Ma, Dyna Robotics”, Oct 13, 2026. Watch the full episode

“What's really hard is that the customers believe or validate that your robot creates enough value that they want to deploy your robots across many of their sites.”
Jason Maat 39:37 · Latent Space39:37

Takeaways

  1. Din Tai Fung expanded Dyna from napkin folding to food cutting after scaling
  2. Deformable objects defeat specialized machines; general models adapt faster
  3. Shared weights give consistent quality across locations via OTA updates
Watch 4:05 from 38:50

In this episode

Full episode

  1. 38:50Multi-site rollouts, not pilots, prove a robot creates value4:05

  2. 56:32Why predicting raw video beats VLAs, and where that bet could fail4:16Up next
  3. 10:12Robot data quality changes as the model improves3:36