In a broad sense, we can think of the economy in terms of resource inputs to society and the dependencies that result in goods and services. This is the kind of

In a broad sense, we can think of the economy in terms of resource inputs to society and the dependencies that result in goods and services. This is the kind of thing you can plot out, industry by industry, with Wardley mapping. Now, clearly, resource extraction jobs aren't going away. The methods change as we go from "ape poking at insects with stick" to "hydraulic fracking", but someone has to have some oversight of the process. But everything leading to a resulting experience fulfilling some human need is subject to change with AI in the picture, because we can imagine machines in the loop doing all the processing, moving, coordinating, and final delivery. Experiences that used to be jobs don't have to be anymore. And therefore the whole social contract of industrial economies is subject to a gradual shift towards indifference: if I have some land, some machines, some access to raw materials, I don't need a job, or to sell things, or to run an organization. But, counterbalancing that, there still have to be voices coordinating and controlling access to those things, because there are too many ethical dangers involved in making it completely permissionless. What financial systems(hence money, wages, etc.) do in controlling access is a way of assigning credit, so that you gain general permissions to stuff. If you do things that persistently assign credit to yourself, without also turning it into a reassignment of credit that makes enemies(power-monging etc.), you survive in any economy. The role of anything artistic in that is, broadly speaking, going to be one of storytelling and aesthetic presentation of information. And I think the key insight to take from the tools we're getting is that direct prompt engineering is going to be taken for granted; but prompt engineering as a layer on top of or beneath a more specialized skillset will still produce distinguished results. That is, if you can glue a little bit of API code together, you can create a job where 90% of your creative time is still spent in a more hands-on mode, but then you send it through the machine as the last step. The trick is in not getting too dependent on someone's platform control play in the process.