Thinking of AI and how it relates to software development, I'm working in the old mode and the new mode. The old mode is I build a project over a few years. I try to bury bits of functionality behind interfaces, either APIs or UIs, and hope I can forget how they work and just access them via the interfaces. Repeat the process. In the new mode, I rely on the machine to remember all that. Claude Code is the key to doing that, using a GitHub repo. And then two or more people can work at the higher level. Obviously the next thing is to see if there aren't some interfaces we can build that are even higher level. The evolution of AI and languages go hand in hand. On the other hand, human beings being what we are, it's just as likely as there will be a wild proliferation of new even more complex interfaces, because now we can rely on the machines to remember the complexities, and their limit is, compared to humans, practically infinite.