Summary
Eric Landau, co-founder and CEO of Encord, joins a Y Combinator talk to make the case that physical AI represents the next major platform shift in technology — and that the data infrastructure layer underpinning it is already a present-day market, not a future bet. Landau traces an unlikely path from particle physics and a decade in high-frequency trading to founding Encord in 2021 during the depths of COVID uncertainty.
The conversation unpacks the ‘bitter lesson’ in AI research — the finding that scaling data and compute consistently outperforms hand-engineered domain expertise — and traces how that principle now drives physical AI development. Encord originally focused on computer vision as the densest and hardest data modality before expanding into full multimodal AI as robotics, autonomous vehicles, and manufacturing applications accelerated sharply after ChatGPT’s 2022 launch.
Landau argues the scale of the physical AI opportunity is frequently underappreciated: 80% of global economic activity involves moving or manipulating things in the physical world, and leading robotics researchers project more robots than people within 5–10 years. For Encord, that market inflection has already arrived, with major robotics and autonomous vehicle teams now among its core customers. The talk offers a useful investor and operator lens on where data labeling, multimodal annotation, and evaluation infrastructure fit in the emerging physical AI stack.
📺 Source: Y Combinator · Published July 25, 2026
🏷️ Format: Interview
