Robotics-at-Scale Pipeline Chart Sparks Debate: Almost All Bitter Lesson
KyleMorgenstein · x · 2026-11-10
Shubham Singh shared a pipeline chart arguing that scaling robotics should follow the bitter lesson almost everywhere: humans set goals/limits/fallbacks, an LLM generates task pairs, checks and rewards; per-task RL specialists are trained in sim and distilled into a generalist; real data corrects sim physics; real-world RL runs on actual robots scored by a learned judge (VLM + human spot-checks). Humans stay in the loop only for safety and judge training. Kyle Morgenstein endorses it as the path to robotics at scale.
More from Embodied
- Apple is building a world-class robotics team, recruiting at CoRL — iamfakhrealam · 2026-11-10
- Using open models to reverse-engineer a Surface Pro Linux kernel build, almost done — vincent_koc · 2026-11-10
- CoRL 2026 to host twin workshops on adapting pretrained robot policies — huihan_liu · 2026-11-10
- Robbyant releases LingBot-Touch, a tactile model giving robot hands human-like touch sensing — CyberRobooo · 2026-11-10
- Dyna Robotics shows RL vs IK head-to-head on wheeled robot, unveils workflow agent Dyna-2.1 — chris_j_paxton · 2026-11-10
- AI researcher Chris Paxton comes away positive from a Tesla Cybercab ride — chris_j_paxton · 2026-11-10