[daxfohl]

Not for a while. They seem decent at generating scaffolding or algorithms, but like for any of the projects I've done recently, it's "figure out what this big piece of mangled code does, take some requirements or high-level idea that's not fully complete and probably has some contradictions, work with people to flesh those out, figure out a way to cram that in, which layer of the code it should be in, or which layers, with some implicit tribal understanding what is allowed to change and what isn't." AI may be able to help with some of that, but it won't be a complete replacement for quite some time.

Also the quarterly planning process is "figure out the ten things you want to get done out of the hundred in your backlog", and the quarterly execution process is "land the one thing you can get done of the ten commitments you made." So there's no lack additional work to be done even if AI takes over some chunk of it.

[muzani]

"figure out what this big piece of mangled code does"

This is my use case for GPT-4. It usually does it better than most humans, including those who wrote the code and those who have maintained the codebase for years.