I don't believe that literal typing of code is the limiting factor in development work. There is the research and planning and figuring out what it is even you need to develop in the first place. By the time you know what questions to even ask an LLM you are not saving much time in my opinion. On top of that you introduce the risk of LLM hallucination when you could have looked it up from a normal web search yourself in slightly more time.
Overall it feels negligible too me in its current state.
I think it depends a lot on the task. While you’re right that just typing is rarely a bottleneck, I would say that derivative implementations often are.
Things like: build a settings system with org, user, and project level settings, and the UI to edit them.
A task like that doesn’t require a lot of thinking and planning, and is well within most developers’ abilities, but it can still take significant time. Maybe you need to create like 10 new files across backend and frontend, choose a couple libraries to help with different aspects, style components for the UI and spend some time getting the UX smooth, make some changes to the webpack config, and so on. None of it is difficult, per se, but it all takes time, and you can run into little problems along the way.
A task like that is like 10-20% planning, and 80-90% going through the motions to implement a lot of unoriginal functionality. In my experience, these kinds of tasks are very common, and the speedup LLMs can bring to them, when prompted well, is pretty dramatic.
> There is the research and planning and figuring out what it is even you need to develop in the first place.
This is where I have found LLMs to be most useful. I have never been able to figure out how to get it to write code that isn't a complete unusable disaster zone. But if you throw your problem at it, it can offer great direction in plain English.
I have decades of research, planning, and figuring things out under my belt, though. That may give me an advantage in guiding it just the right way, whereas the junior might not be able to get anything practical from it, and thus that might explain their focus on code generation instead?
Overall it feels negligible too me in its current state.