Daimon Robotics targets 1 million hours of tactile data

Daimon Robotics' strategy stands out because it deliberately avoids competing with the companies building general-purpose robotic brains. Instead, it aims to supply the underlying tools and data infrastructure that those developers can use. The company has identified Lenovo Capital and Inovance Technology as key partners and says it has moved beyond proof-of-concept projects into industrial assembly lines and warehouses, though it declined to name specific customers due to nondisclosure agreements.

Collecting tactile data at scale is a different problem from collecting visual data. Tactile information depends on physical contact and varies with an object's material, geometry, the forces applied, and the task, which makes it harder to capture with cameras. Daimon's pipeline uses tactile feedback operating at 30 Hz and works to calibrate simulations against real-world measurements before transferring data into virtual environments. Dr. Zhu was candid about the current state of simulation, calling it "kind of terrible" when describing the gap between simulated and real-world environments.

Even with large datasets, the models remain limited. Zhu said the model "is not as intelligent as we think" and that its generalization ability is sometimes in-domain, meaning it may struggle to transfer to unfamiliar tasks. Daimon's success will depend on whether its approach measurably improves manipulation, reduces the effort required to develop new applications, and supports deployments across different industrial environments.