Optimus

Tesla showcases new Optimus capabilities, claiming its humanoid robot can now learn complex tasks from human video demonstrations, aiming to accelerate skill acquisition. This comes as the field of robotic reinforcement learning sees rapid advancements from multiple players.

Tesla provided updates on its Optimus humanoid robot, detailing ambitious production targets including a future 1,000/month line (still months away), and eventual goals of 100,000/month. CEO Elon Musk also stated the current Gen 2 Optimus is 'far from its final form,' as the company hires data collectors to refine its AI, which uses larger neural nets than its cars. Commercialization is eyed for mid-2026 with a $20k target cost.

Recent Tesla Optimus videos have sparked discussion around robot foot design, specifically the role of articulated and actuated toes in enhancing humanoid agility and efficiency. A closer look reveals different approaches across the industry.

Tesla's Optimus robot demonstrates more fluid, human-like movements, touted by Elon Musk as 'real real-time.' The update highlights rapid progress in humanoid robotics, driven by simulation and reinforcement learning, a trend also visible in recent demonstrations by competitors like Unitree and EngineAI, and academic research projects such as HuB.

Soft Robotics Podcast host Marwa ElDiwiny and engineer Scott Walter dissect the rotary-vs-linear actuator debate shaping today’s humanoid robots—explaining why most builders favor compact rotary joints while others bet on planetary-roller-screw cylinders.