Humanoid Robotics Timelines Accelerate by Three Years, Says Former OpenAI Researcher

San Francisco, CA – Rohan Pandey, a prominent AI researcher and co-founder of Periodic Labs, has publicly stated that the development timelines for humanoid robotics are advancing approximately three years faster than his previous expectations. The announcement, made on social media on October 29, 2025, highlights a significant shift in expert perception regarding the pace of progress in embodied AI.

Pandey, formerly of OpenAI, expressed his revised outlook, tweeting, > "humanoid timelines are running ~3 years faster than i expected as of yesterday." This statement comes amidst a flurry of recent advancements and significant investment in the humanoid robotics sector. His current work at Periodic Labs focuses on autonomous science, aiming to accelerate AI through experimental verification, giving him a unique vantage point on technological progression.

The accelerated timeline perception may be influenced by recent industry developments, including a bullish assessment from Ark Invest CEO Cathie Wood. Speaking at an investment conference on October 28, Wood dismissed "AI hype" concerns, asserting that humanoid robots represent the "biggest of all the embodied AI opportunities" and predicting transformative impacts across various sectors. Her comments underscore a growing investor confidence in the commercial viability and rapid deployment of these advanced machines.

Recent product launches and demonstrations further illustrate the rapid progress. Figure AI, for example, unveiled its Helix AI model and the Figure 03 humanoid, with plans to test in homes two years ahead of schedule. Similarly, UBTECH showcased an "intelligent swarm" of its industrial humanoids and is preparing to deliver its Walker S2 models, while Unitree introduced the agile H2 and D-infinite by Direct Drive Tech. These developments suggest a maturing ecosystem for humanoid robot capabilities.

Despite the optimism, not all experts share this accelerated view. Yann LeCun, Meta's Chief AI Scientist, recently cautioned that most robotics companies are overly focused on hardware, lacking the necessary intelligence to make humanoid robots truly useful. LeCun emphasized the need for AI systems to learn from high-bandwidth sensory data like video, suggesting that current large language models are insufficient for powering advanced humanoid functions. This ongoing debate highlights the complex challenges and varying perspectives within the rapidly evolving field of humanoid robotics.