Robot Startup Ditches VLAs for Frontier LLMs in a Customer Deployment
ihorbeaver · x · 2026-10-16
A robotics company switched a customer deployment entirely to frontier LLMs controlling its robot, building a "System 1/System 2 on steroids" architecture:
- High-level reasoning: an agent based on Opus and Astra, which the author says are many generations more capable than any VLA
- Fast execution: computer vision code written by the model—fast, editable, and reliable
Bonus benefits: in-context learning as a baseline capability, and continual learning—the large model steps in and adapts the pipeline whenever new situations arise. The author frames this as being on the right side of the Bitter Lesson, riding the billions invested in frontier models worldwide.
More from Embodied
- Dev demos open wearable AI passport running multilingual flashcards on tiny memory — ezshine · 2026-10-16
- Robots enter the tarmac at Japan's busiest airport for a roughly three-year live test — Aiden_Tech_Ai · 2026-10-16
- Team says no technical blockers remain for v1 self-replicating factory using off-the-shelf parts — ihorbeaver · 2026-10-16
- Skip the fancy vision: a $300 automatic screw feeder beats LLM-controlled picking — ihorbeaver · 2026-10-16
- After 50+ robotaxi rides, user ranks Tesla Cybercab the best rider experience — EricETesla · 2026-10-16
- CMU's Andrea Bajcsy on what safety means for generalist robots, beyond collision avoidance — CSProfKGD · 2026-10-16