The Robot Solution and the Next Step in Programming

We are trying to mainly go for a skill-based approach. So it's not a task of the reinforcement learning to make everything but to solve one specific task. And then you have a skill in the end, you have a capability that you can use and plug together with other things. How difficult is it to get those worlds together? Because it's a totally different approach, I think. But there we have some really promising results. If we assume having this robot as a machine to build this so-called task graph where you have tasks and you can even move dynamically on the task graph.