The Future of Humanoid Robots With 1X's Bernt Bornich - Ep. 259

What It Covers

Bernt Bornich, CEO of 1X Technologies, explains how his company builds safe, affordable humanoid robots using tendon-driven systems that learn through reinforcement learning, teleoperation data, and real-world failure to enable autonomous home assistance.

Key Questions Answered

• •Safety Through Low-Energy Design: 1X's tendon-driven robots weigh only 30 kilos but lift 150 pounds by using low gear ratios (0.7-1.2:1 versus 100:1 in industrial robots), reducing internal energy by 10x and enabling safe human interaction and learning through failure.
• •Learning From Failure Strategy: Robots achieve initial task success through teleoperation demonstrations and internet data, then autonomously improve by attempting tasks repeatedly, distinguishing successful attempts from failures, creating a self-improving data flywheel that scales beyond human teleoperation limits.
• •World Models Enable Prediction: World models allow robots to simulate future outcomes based on current state and planned actions, creating probability trees that enable backward search from goals and dramatically improve data efficiency for learning physics and complex manipulation tasks.
• •Consumer Launch Timeline: 1X ships Neo humanoid robots to consumer homes in 2025, capable of vacuuming, tidying, doing laundry, and folding clothes, positioning the product as a developmental journey where customers participate in teaching robots through daily interaction, not a finished solution.