The Secret Supplier Feeding Frontier AI

Grass serves as a decentralized data infrastructure platform that enables AI models to access the public internet at exabyte scale. As websites increasingly block data center IP addresses, Grass utilizes a network of millions of opt-in residential nodes to bypass these restrictions, providing ethical, consented data for AI training and inference. CEO Andrej Radonjic highlights the shift from static training data to live context retrieval, where agents require real-time, up-to-date information to remain functional. By owning its infrastructure—ranging from compute servers to storage—Grass maintains a competitive moat, allowing it to provide high-quality, targeted data sets for diverse applications, including robotics and generalist AI models. This infrastructure-first approach ensures that AI systems can bypass anti-bot measures while maintaining the integrity and diversity of data necessary for model performance and continuous learning.