Historically, many master painters used teams of assistants/apprentices to do most of the work under their guidance, with them only stepping in to do actual painting in the final details.

Similar with famous architects running large studios, mostly taking on a higher level conceptual role for any commissions they're involved in.

Traditionally in software (20+ years ago) architects typically wouldn't code much outside of POC work, they just worked with systems engineers and generated a ton of UML to be disseminated. So if we go back to that type of role, it somewhat fits in with agentic software dev.

Sure, but you currently cannot teach AI models to generate novel art in the same way that you can teach a human apprentice.

That's where we're at a marked disagreement. "It's just a way to get paid" reduces every human knowledge to a monetary transaction, so the value of any type of learning is only worth what is being paid for.

Thankfully the people that came before us didn't see it that way otherwise we wouldn't even have anything to program on.

And how did the master painter learn his craft, without first having been an assistant or apprentice?

> they just worked with systems engineers and generated a ton of UML to be disseminated. So if we go back to that type of role, it somewhat fits in with agentic software dev.

I've never met one of those UML slingers that added much value.

You can still enjoy painting, but there is no guarantee that you will be paid for it.