yeah maybe I could finish the functions in glm and then let Gemini to review the logic and make a nice design. That is a great idea and a good test. After that I could let grok create a summary and try it again with both models (: