Two ways to run GPT-6 Astra on robots: direct VLA policy or agent with tools
YuXiang_IRVL · x · 2026-09-17
Researcher YuXiangIRVL outlines two distinct ways GPT-6 Astra is being used for robot manipulation:
- As a robot policy (VLA-style): Astra directly outputs robot actions end-to-end.
- As an agent: Astra uses tools, code, planners, or other policies to control the robot indirectly, acting as a high-level decision maker.
Demo examples of both approaches are included. The distinction maps onto the core debate in embodied AI: end-to-end VLA policies versus layered agent architectures.
More from Embodied
- Scoble hails Snap Specs as 'way ahead of anyone else' in AR glasses — Scobleizer · 2026-09-17
- Blogger predicts Figure's teased breakthrough is robot speed-learning its next job — imjustnewatai · 2026-09-17
- NVIDIA's SOMA Retargeter update adds Robot Configurator for humanoid motion transfer — RubenEVillegas · 2026-09-17
- Embodied AI hasn't escaped gravity: humanoid demos repeat cobot boom failures from a decade ago — yongqianme · 2026-09-17
- Deep dive: what's really happening behind the Tesla Optimus, Figure and 1X NEO humanoid race — MickeySteamboat · 2026-09-17
- Figure teases an 'AI breakthrough' with a demo set for tomorrow — adcock_brett · 2026-09-17