Ex-Tesla Optimus engineer: scaling human data key to robotics, physical AGI needs more
mihdalal · x · 2026-09-16
An engineer who recently left Tesla's Optimus team reflects on his work spanning low-latency visuomotor systems, on-device model deployment, sim-to-real whole-body control and dexterity, and scaling imitation learning on large human data. He remains very bullish on scaling for robotics—especially human data—but argues physical AGI still requires fundamental breakthroughs in video models, scaling laws and mastery. He'll announce his next role soon.
More from Companies & People
- AI roundup: Gemini 3.8 Live for voice agents, Huang pushes back on regulation — rohanpaul_ai · 2026-09-16
- Zuckerberg: Meta delayed Muse for months over safety, in jab at Anthropic's slowdown call — soumitrashukla9 · 2026-09-16
- AI Conference 2026 in SF nearly sold out: 120+ speakers incl. Norvig, Stoica, Lattner — bigdata · 2026-09-16
- NYT: Zuckerberg jabs Anthropic in fight over AI slowdown, says safety over scaling — nordicinst · 2026-09-16
- Anthropic hires Sihao Huang as Head of Frontier Compute Strategy under Tom Brown — NotTomBrown · 2026-09-16
- Gary Marcus: synchronized AI slowdown protects trillion-dollar market values — GaryMarcus · 2026-09-16