> I don't disagree with you that LLM's remove the need for type-safe languages

That feels like a really strong conclusion. I’m not clear on why any safeguard isn’t a useful safeguard if you let agents write all the code.