I honestly feel like you're making this sound more complicated than it needs to be. I get what I would describe as very good results from GPT 5.6 on my projects.
There are some methodologies that can improve things for me versus just YOLO'ing but even these are of marginal benefit:
* Have good requirements. Experience with a codebase and stakeholders helps a lot here.
* Correctly subdivide the task into chunks that won't blow context. You can write a big task and have an agent plan subtask delegation for you, but it's good to have some intuition of your own.
* Perform an automated code review. This is a no-brainer but it catches stuff.
* Make sure you understand the "big picture" stuff and stop caring about the little details. The agents will write unit tests, so you shouldn't have to care about reading every LOC, you can ask the agent to describe the architecture and flow instead.
This is still weird to me, the agents are super-good and clever most of the time, but I do feel I always need to direct them to a small area to focus: much like a human!! If you just ask them to implement things, they never (for me anyway, were not allowed the most expensive model! Terra is it for now) suggest they should stop adding code ontop of code and refacor, I always have to poke them to do that. Having done that once, and added some tests, they suddenly become aware that, yeah, maybe we should test stuff.
The LLMs seem to have no innate ability to understand whats a good direction a higher level. I mean, if you ask them about it, they will actually kinda figure that out, too. But always need that nudge...
So if you as a developer do not have the innate drive to ensure quality, the results will be terrible in my experience.
If you DO spend the tokens on quality though, it can also be kinda awesome. But its not magic.. I notice clear "slowdowns" the bigger the scope gets. They are not actually able to, in any way, subdivide implementations more efficiently than humans.
There are some methodologies that can improve things for me versus just YOLO'ing but even these are of marginal benefit:
* Have good requirements. Experience with a codebase and stakeholders helps a lot here.
* Correctly subdivide the task into chunks that won't blow context. You can write a big task and have an agent plan subtask delegation for you, but it's good to have some intuition of your own.
* Perform an automated code review. This is a no-brainer but it catches stuff.
* Make sure you understand the "big picture" stuff and stop caring about the little details. The agents will write unit tests, so you shouldn't have to care about reading every LOC, you can ask the agent to describe the architecture and flow instead.