Jelly Sheet (mrdoob.github.io)

That said, I use local AI on GPUs I own, so they are -my- tools on -my- property to help me research debug and help me execute my ideas faster than I could before, so more things get done than were done previously.

I keep my AI rack under my 3D printer because they are the same type of tool to me.

If you want to draw out the construction analogy, ask a mason or carpenter if they'd rather be an architect. (a lot would probably not want to, but I know a few people in the trades that can do a lot more than just welding two pieces of steel together, it's just that their job is the welding part.

Actually this will be one of the challenges; LLM assisted coding has empowered people to be more decision makers, lead developers, architects, etc than code monkeys. But it's a pyramid, there's only so many lead developers you feasibly need on a project.)

Yes. Writing code was always fun but running a project is itself a systems engineering problem. Especially if you’re the one defining project scope.

So yes, absolutely it is the same level of joy. The code I wrote was just the method needed to solve the low-level problem but it was never the entire problem space.

I can see my fingerprints all over the user facing design and I can see all the places where I agonized over different approaches to maximize user enjoyment. Seeing the solutions manifest is thrilling frankly. I see users of LLMs as filling a similar role to mine in this respect.

That said, I understand that not everyone enjoys the entirety of the problem space to the same degree and so I can see how a reasonable group of people might not find the higher-level puzzle as enjoyable as its low-level counterpart just as there are those who might find the low-level tedious instead of rewarding.