You may be right. For my part I don't agree. I think the scale will be so much greater than the jump from 1995 to 2020. The moat will be data, or anything derived from hardware/the physical world. That is the tie back into "the ability to clone is not sufficient". You can clone Amazon app but how will you ship a package? Especially as their dominance in that area rests on their monopoly power?

Eventually they will fall too. It will just take longer. I think really the reason I disagree is simply that AI will get good enough where necessarily no human will have an edge. The super advanced ones will be on the whole time ambiently and likely adversarially self prompting. Your great idea already exists somewhere out in latent space. They will either get their first or just copy you. You will not care about anyone's edge. When you care about something you will necessarily engage with it manually. I mean this in the definitional sense. How did we know Alice cares about task x? Because she did it manually. She doesn't really care about the weather, so she just asked the computer in enough ways to make her own version of that. In the same way you wouldn't think Alice cared about Bob if she delegated correspondence to him to someone else. For everything else compute as utility will have you covered.