{% extends "base.html" %} {% block title %}GameGame minds and curriculum{% endblock %} {% block page_id %}gamegame_mind{% endblock %} {% block head %}{% endblock %} {% block content %}
Native learned JSON, measured play records, actual model creation and the shared circuit lesson list.
Create uses the desktop's exact Modelfile generator, including the saved learned lessons. It creates a real tag on the selected model target; it does not apply that tag automatically. Failed or interrupted model creation can leave a target artifact: the actual result reports what was confirmed.
Try and Keep select the saved job's tag in the native profile and apply it to the actual selected target. This can restart the Bot and resume its configured game activity. Compare reads the running Bot's measured record; it does not run games or invent a score. Keep uses the desktop's promotion gate: at least 20 challenger episodes and a win rate that beats the recorded baseline, when a baseline exists.
Save edits the installed Bot source's curriculum.json, preserving the native object or array format and unknown fields. A curriculum in the Bot's own data folder takes precedence. This save does not deploy the source or change a running Bot's loaded curriculum; use the native controls to send or restart it.