I’m more comfortable with LLMs if I think of them as autocomplete for thinking. They don’t reason and aren’t really AI but they can offer decent suggestions. And maybe that’s ok?

The challenge, as with all software, is doing it privately and safely.