Ones that can remediate it though. If I am capable of safely refactoring 1,000 copies of a method, in a codebase that humans don’t look at, did it really matter if the workload functions as designed?

Jeebus, 'safely' is carrying a hell of a lot of water there...

In a type safe language like C# or Java, why could you need an LLM for that? it’s a standard guaranteed safe (as long as you aren’t using reflection) refactor with ReSharper.

Features present in all IDEs over the last 5 years or so are better and more verifiably correct for this task than probabilistic text generators.