I think "reliable answers" goes a bit against LLM philosophy lol

Definitely, the entire orchestration "intelligence" is very different, and I think we have yet to see how security and testing will be handled... Very interesting times!!!

Exactly, the orchestration is handled in a very different way. I'd assume QA would need to be handled by separate agents too... I played with AutoGen and it's much easier to configure then an AppMesh for micro services:)

That's it... You don't "program them", you explain to them in English what they are, where they read and write the information... But they do the same thing as hardcoded dockers