> With Managed Agents, you define outcomes and success criteria, and Claude self-evaluates and iterates until it gets there (available in research preview, requ

> With Managed Agents, you define outcomes and success criteria, and Claude self-evaluates and iterates until it gets there (available in research preview, request access here). It also supports traditional prompt-and-response workflows when you want tighter control. Call me stupid, but this sounds not like they want software developers to be around in a year or two.