Since I am currently using AI to make a game engine, I feel like I am sort of qualified to say something on the topic, and I don’t think it’s really true that AI’s can just roll one out without a lot of poking. They constantly go off on the wrong track and I have no faith that they’d get anywhere sensible without a human steering the whole thing (for now anyway).

But even if they could the best thing they’re ever going to do is reach the sum total of human knowledge. I can’t ask it to think of a novel algorithm to speed up my flows, and even the places where I feed it a literal paper and ask it to reproduce it in a production ready matter are often exercises in patience.

That said, for me it’s better because I like thinking of ideas far better than typing out code for ideas to become reality.

Yep. AI can probably make your marketing website or REST service on autopilot. Or even small minigame experiences. But go outside the bounds of median programming and AI becomes increasingly scatterbrained. It requires a lot of guidance to successfully build high performance, novel software. I still find it more productive though - I can work on three tasks simultaneously with AI, and only one if I’m manually working on a single algorithm. It’s very much like running a team of well-read junior developers.