Robotics' most-hyped model completed just 7 of 100 real manipulation tasks, demo reel hides the rest
carrycooldude · x · 2026-10-05
boundless VC notes the most hyped robotics model completed only 7 of 100 real manipulation tasks — removing a marker cap, passing a ruler, opening a box — with the demo reel skipping the other 93. The author argues the real job of deploying physical AI is evals that catch failures, failure-data collection, and the retraining loop. boundless and Qualcomm will co-host a closed technical roundtable on post-deployment physical AI in Bangalore on Oct 22.
More from Embodied
- Dream4ACT Unifies Video-Action Modeling Across Robot Embodiments, Hitting 89% on RoboTwin 2.0 — Xiangyu Zhu · 2026-10-05
- Munich Robotics Startup RobCo Hits $1B Valuation in Nine Months — lukas_m_ziegler · 2026-10-05
- Physical AI field session in Bengaluru to tackle post-deployment evals and retraining loops — carrycooldude · 2026-10-05
- Meta's NAVA-WAM pretrains robot action policies directly from action-free videos — meta · 2026-10-05
- Sergey Levine on robotics: hardware is good enough, the real gap is decision-making and data — 机器之心 · 2026-10-05
- SJTU's LIFT adds force sensing to VLAs with zero force-labeled pretraining data — jiqizhixin · 2026-10-05