Looking at the quoted text one could guess his goal is maximized amount of users, alternative maximized number of copy-paste-modify events, and that he thought

Looking at the quoted text one could guess his goal is maximized amount of users, alternative maximized number of copy-paste-modify events, and that he thought that the GPL would produce those? (We could dig into the terminology around "openess", a classic tar trap in the GPL and BSD flame wars, but I doubt such path would lead towards comments getting more thoughtful and substantive.) Going thus back towards getting users and copy-paste-modify events, as a strategy GPL is rarely if ever a maximize for those goals. The discussions around licenses for stackoverflow code seems to have concluded that claiming copyright at all on small code snippets is counter productive and something most users don't seem to even acknowledge or know that it has a license. A patent grant is however a good protective measure, which would be more akin to the apache license rather than BSD. For maximum users we have both Iphone and Android as examples of BSD vs GPL, and the conclusion seems to be that the license has very little effect. In contrast one could make a strong correlative argument that proprietary software on top and DRM is the best strategy out there if you want to maximize users. I doubt however either is what Carmack is talking about. Instead it seems to be the rather common pattern we have seen in the past when old developers test out open source. They look at projects like linux and expect to get similar contributors by releasing their own projects under GPL, and then they end up not getting many and announcing that "the best aspects of GPL work didn't manifest". The question is if that would also have happened under BSD. He could try it today, and if we got the same result, what would that prove?