Thanks, it's an interesting perspective. I may have underappreciated the code volume/time requirements to execute especially in a startup setting.

I think I see where you're coming from. I'd imagine in some applications, there is very niche/specific implementation on Github that is both expensive to reimplement and not permissively licensed. I can see the "AI laundering" angle there.