Fine-tune robot foundation models safely.

"You're thinking about the wrong risk. I'm not worried the labs will compete with me directly. I'm worried they'll commoditize the data advantage I spent years building. Once that happens, every startup in my space gets access to the same capabilities, and suddenly I'm competing with a dozen fast followers instead of protecting a hard-won lead."

"We realized there's a fundamental IP risk in handing our model directly to a robotics company. To make the system useful, we would need to provide model weights, inference code, or fine-tuning capabilities. At that point, the customer effectively has permanent access to core parts of our technology, even if the partnership ends. From our perspective, that creates a difficult tradeoff between enabling deployment and protecting the long-term value of our model IP."