My take is that if the LLM outputs text for humans to read, that's not an agent. If it's making API calls and doing things with the results, that's an agent. Bu

My take is that if the LLM outputs text for humans to read, that's not an agent. If it's making API calls and doing things with the results, that's an agent. But given the way "AI" has stretched to become the new "radium" [1], I'm sure "agent" will shortly become almost meaningless. [1] https://en.wikipedia.org/wiki/Radium_fad https://en.wikipedia.org/wiki/Radium_fad