Are there places where it reliably saves time, though? Unlike most software tools, LLMs are stochastic, right?
In my experience they reliably waste much more time than they save