Perhaps two agents that write and read from the same memory store frequently throughout their generation. More specific than that I couldn’t say

So it’s more in choosing the specific model for the toolset? Could this not be considered part of the “mode”?

If I give an LLM the context A and tools B in one mode and then give it context C and tools D in another mode, how is that different to two agents one with context A tools B and one with context C tools D? The difference feels like it’s in the orchestration of deciding which agents to call

I’d put money on a substantial amount of production use cases requiring structured generation. If Anthropic want to be the big player in the API space they need to get a move on

It feels equivalent except in the scenario where there is a benefit to multiple modes collaborating in parallel simultaneously. Perhaps if token streaming could be interrupted by a message from another agent it would be more useful, but I don’t think the big players support this