Skill building, sharing and model fine tuning

[vmandrilly]

Hi! From the little I have seen from the project, it tries to improve on open interpreter to try to build a skill base that can be reused. That's nice, but at first glance it looks like trying to patch a leaking pot: yes, for the specific use, you are going to be able to reproduce the good behavior, but the original AI model has not been improved a bit. What you really want is that the AI model (whichever you are using) is improving and is giving you consistent actions according to your computer environment. For this, you need to build a dataset from all users actions and the feedback (user feedback if user gives feedback, or computer feedback - success or error to run the script) and associated user computer environment, so that the system can fine tune the AI model you are using.