You need to be able to iterate and extend the codebase, as well as grow it to 100K+.

Personally, I would focus everything on making changes easy to make at large scale. Typically with a human programmer this comes from designing and architecting and coming up with clever solutions that allow new features to be built faster even with a lot mroe lines of code. Maybe AI can do that, but I think it's uncharted territory because the biggest companies use AI with established codebases.

Trash bin? ~1M LOC for what? If you cannot review or test it, then why would you trust it?

This is why it's better to stay human in the loop during the process. Sounds like you don't think it should take 1M LOC. It's likely your garden variety slop, even if it "works". Avoid the sunk cost fallacy, Ai lets us throw away more experiments because they are "cheap".