The real bottleneck in robotics isn't hardware or compute - it's the absence of physical-world data. Decades of dexterity, force, and intuition live only in how humans operate in the physical world, which has never been captured. In this talk, Rushil Agarwal will share how Human Archive is changing that by capturing and synchronizing video, tactile, audio, and motion data from over 100,000 contributors doing real work across homes, restaurants, factories, hotels etc. , then structuring it into something machines can learn from.

Rushil will walk through the founder's journey behind it - why they bet on human data over teleoperation, the hardest problems they hit scaling hardware and operations across the globe, and the unexpected challenge of building an entirely new category, where a blue-collar worker has never been asked to generate training data before. He will close with an honest look at where the industry actually stands: how close the "GPT moment" for robotics really is, what's still unsolved, and what it unlocks for the future of work, cognition, and beyond.